fix(js-instrument): match native function arity and name on wrappers

[vringar]

What

The legacy JavaScript instrument (Extension/src/lib/js-instruments.ts) replaces
native methods with a closure wrapper (instrumentFunction). The wrapper's
page-observable length (arity) and name did not match the native
function it wraps: every wrapped method reported length === 0 and name === "",
regardless of the native value. Both are trivially readable from page script and
make instrumented methods fingerprintable.

This PR copies the native function's own length and name property descriptors
onto the wrapper (preserving the native non-writable / non-enumerable /
configurable shape), so the wrapper matches native on these two axes.

Empirical before/after (direct-selenium, FF152, no extension)

Repro injects the real compiled getInstrumentJS (via .toString(), exactly
as the extension's ${getInstrumentJS} template-literal does) as an inline
<script>, headless, no extension. It instruments concrete native methods and
reads the wrapper's .length/.name:

method	property	native	wrapped (before)	wrapped (after)
CanvasRenderingContext2D.prototype.getImageData	.length	4	0	4
CanvasRenderingContext2D.prototype.getImageData	.name	"getImageData"	""	"getImageData"
HTMLCanvasElement.prototype.toDataURL	.name	"toDataURL"	""	"toDataURL"

Both artifacts reproduced before the fix and are gone after. Instrumentation
still logs the calls in both runs (no capture regression).

Regression test

test/test_js_instrument.py::TestJSInstrumentFunctionArityAndName instruments
window.fetch in a real browser, reads the wrapper's observable .length/.name
in-page, and encodes them into a fetch call argument so they land in the
javascript table. Native fetch has arity 1 and name "fetch"; the test
asserts the logged call argument is https://arity-1.example.com/name-fetch.

Verified the negative: with the fix disabled (wrapper left unchanged) the test
fails (observed arity-0 / name-), confirming it is a genuine guard. The
full test/test_js_instrument.py suite passes (8 tests).

Out of scope

The separate proto-chain flattening artifact (#56 sub-item iii) is not
addressed here. For the legacy plain-JS-wrapper architecture, de-polluting the
flatten would silently drop inherited-member capture that the bundled
collection_fingerprinting relies on (per-interface attribution of inherited
members is structurally tied to the per-leaf copy-down). It is tradeoff-blocked
and intentionally left untouched.

Refs crosslink #56.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 62.20%. Comparing base (61d285f) to head (7bef7e2).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1210      +/-   ##
==========================================
- Coverage   62.22%   62.20%   -0.03%     
==========================================
  Files          40       40              
  Lines        3929     3929              
==========================================
- Hits         2445     2444       -1     
- Misses       1484     1485       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.