Reset the embed update log when it cannot be read

[pasmud]

Thanks for contributing, make sure you address all the checklists (for details on how see development documentation)

• ran the linter to address style issues (tox -e fix)
• wrote descriptive pull request text
• ensured there are test(s) validating the fix
• added news fragment in docs/changelog folder
• updated/extended the documentation
• for changes to activation scripts, pyvenv.cfg, wheel downloads, app-data or release workflows: explained in the
  PR text which outside input the change handles, such as a path, a prompt or a downloaded file, and how it stops that
  input from running as code. See the
  security scope.

The bug

The embed update log lives at <app-data>/wheel/<major.minor>/embed/3/<distribution>.json and records which newer
seed wheels are known. UpdateLog.from_app_data() in src/virtualenv/seed/wheels/periodic_update.py passed whatever it
read straight into from_dict(), which calls .get() on the result. A value of the wrong shape raises, and the seed
fails with a message that names neither the cause nor a way forward:

fail
Traceback (most recent call last):
  ...
  File ".../periodic_update.py", line 60, in periodic_update
    u_log = UpdateLog.from_app_data(app_data, distribution, for_py_version)
  File ".../periodic_update.py", line 196, in from_app_data
    return cls.from_dict(raw_json)
  File ".../periodic_update.py", line 187, in from_dict
    load_datetime(dictionary.get("started")),
                  ^^^^^^^^^^^^^^
AttributeError: 'list' object has no attribute 'get'
AttributeError("'list' object has no attribute 'get'")
RuntimeError: seed failed due to failing to download wheels pip

Five shapes all do it, verified by calling periodic_update() directly:

list                 -> RAISED AttributeError: 'list' object has no attribute 'get'
string               -> RAISED AttributeError: 'str' object has no attribute 'get'
versions-as-string   -> RAISED TypeError: string indices must be integers, not 'str'
completed-as-int     -> RAISED TypeError: strptime() argument 1 must be str, not int
versions-junk        -> RAISED TypeError: string indices must be integers, not 'str'

End to end, virtualenv <dest> --no-periodic-update exits 1 for each of those and does not create the environment.

This matters because the app data folder is shared and persists between runs. A log written by a different virtualenv
version, or left half written by an interrupted one, is enough to make every later seed fail until the user works out
that they need to delete a cache file.

The change

from_app_data() catches the shape errors, logs a warning naming the distribution, removes the file and returns an
empty log. The bundled wheel is then used and the environment is created, and the next run starts without the bad
log instead of warning again.

This matches what the surrounding code already does: JSONStoreDisk.read() deletes the file and returns None when
the JSON does not parse, so a corrupt log was already survivable, only a log that parses into the wrong shape was not.

Tests

Five cases, all failing on unmodified main (288f13c) and passing with the fix.

$ tox -e py -- tests/unit/seed/wheels/test_periodic_update.py -p no:randomly -k malformed

Before, with src/virtualenv/seed/wheels/periodic_update.py reverted to main:

E       AttributeError: 'list' object has no attribute 'get'
E       AttributeError: 'str' object has no attribute 'get'
E       TypeError: string indices must be integers, not 'str'
E       TypeError: string indices must be integers, not 'str'
E       TypeError: strptime() argument 1 must be str, not int
FAILED tests/unit/seed/wheels/test_periodic_update.py::test_periodic_update_drops_malformed_log[not-a-mapping]
FAILED tests/unit/seed/wheels/test_periodic_update.py::test_periodic_update_drops_malformed_log[string]
FAILED tests/unit/seed/wheels/test_periodic_update.py::test_periodic_update_drops_malformed_log[versions-not-a-list]
FAILED tests/unit/seed/wheels/test_periodic_update.py::test_periodic_update_drops_malformed_log[version-not-a-mapping]
FAILED tests/unit/seed/wheels/test_periodic_update.py::test_periodic_update_drops_malformed_log[completed-not-a-datetime]
======================= 5 failed, 35 deselected in 1.08s =======================

After:

======================= 5 passed, 35 deselected in 0.64s =======================

Each case writes the bad log into a temporary app data folder and asserts the bundled wheel is returned, the exact
warning is logged and the file is gone.

End to end, with a wrong-shape log planted in the app data folder, virtualenv <dest> --no-periodic-update now exits 0
for all four shapes I tried where it previously exited 1.

Wider selections, all passing:

tox -e py -- tests/unit/seed/ -p no:randomly    165 passed
tox -e py                                      474 passed, 50 skipped, 11 failed

Linter and type checks:

tox -e fix        -> OK, all 16 hooks passed
tox -e type       -> All checks passed!
tox -e type-3.9   -> All checks passed!

Full suite

The 11 failures above are unchanged by this branch. I compared the failing test IDs before and after by stashing the
change and re-running:

$ diff <failing IDs on main> <failing IDs on this branch>
IDENTICAL: same set of failures
before: 11  after: 11

They are artifacts of this sandbox, all reproducible on unmodified main:

• tests/unit/test_util.py (2): the safe_delete tests rely on directory permissions denying an unlink, which does
  not happen for uid 0. This container runs as root.
• tests/unit/activation/test_bash.py (9): bash renders the escaped \$ prompt sequence as # for uid 0 and as $
  for any other user, so the expected prompt text does not match. Confirmed by running the same PS1 as root and as
  uid 65534: root gives (a%n#HOME\b...), a normal user gives (a%n$HOME\b...).

fish, csh and tcsh are installed here, so those activator tests ran. powershell and nu are not installed, so
the PowerShell and Nushell activator tests skip. CI runs them, so a green local run does not guarantee a green CI run
for those files.

Security scope

The outside input is the embed update log file, which sits in the shared app data folder. It is data read with
json.loads, and nothing on this path executes or evaluates it. Before, a value of an unexpected shape propagated an
exception out of the seed; now it is reported and discarded, which narrows what a bad file can do rather than widening
it. No new input is read, and the fallback only ever selects the bundled wheel, which is the same one used when no log
exists at all.