ci: align Apple signing secrets with the org-provided names, and stop signing the weekly build

[Naruto]

Summary

Two related changes to how this repo handles the Apple Developer ID certificate.

1. release.yml — rename the secrets. They were provisioned at the organization level under different names than the workflow assumed. This mattered more than a rename usually does: every signing step treats an empty secret as "not configured", skips its work, and returns success. With the names mismatched, a fully green release run would still have produced unsigned macOS/iOS GDExtension frameworks.

Old	New
SS_APPLE_CERTIFICATE	APPLE_DEV_ID_APP_CERT
SS_APPLE_CERTIFICATE_PASSWORD	APPLE_DEV_ID_APP_CERT_PASS
SS_APPLE_SIGNING_IDENTITY	APPLE_DEV_ID_APP_NAME
SS_APPLE_API_ISSUER	ASC_API_ISSUER_ID
SS_APPLE_API_KEY	ASC_API_KEY_ID
SS_APPLE_API_KEY_BASE64	ASC_API_KEY_PASS

Only the secrets.* right-hand sides change. The env var names the release scripts read (APPLE_SIGNING_IDENTITY, APPLE_API_ISSUER, APPLE_API_KEY, APPLE_API_KEY_PATH) are untouched, so release-gdextension-macos.sh and release-gdextension-ios.sh keep working unchanged locally. The script comments that named the old secrets are updated to match.

2. weekly.yml — stop signing entirely. The weekly build is continuous verification of our own tree, not something we distribute, so it has no business handling the Developer ID certificate. Signing belongs to release.yml alone, and keeping the certificate out of a scheduled job that runs unattended every week is the smaller attack surface. This drops the "Set up signing keychain" step, the APPLE_SIGNING_IDENTITY hand-off to the macOS/iOS build steps, and the now-unused KEYCHAIN_PATH workflow env.

The builds themselves are unchanged — the scripts already no-op on signing when APPLE_SIGNING_IDENTITY is absent, so the artifacts are simply unsigned. That was already the tolerated outcome: the keychain step carried continue-on-error: true with the comment 「Weeklyでは署名できなくても通す」.

Notes on the mapping

• ASC_API_KEY_PASS holds the base64 of the .p8 private key, despite what its name suggests, so it maps to the key file the notarization step decodes — not to a passphrase. Worth confirming on the first signed run.
• TMP_KC_PASS is deliberately unused. release.yml still generates the temporary keychain password per run with openssl rand, which is strictly better than sharing one value across runs. A comment records why.
• The sibling repos got the identical rename: SpriteStudio-SDK and SSConverterGUI (both merged).

Verification

Both workflows parse as valid YAML. No SS_APPLE_* references remain in the tree, and after the second commit weekly.yml has no secrets., KEYCHAIN or codesign references left at all.

The signing itself cannot be verified from this branch — release only runs on workflow_dispatch, and org secrets are not exposed to a non-permitted repository. Two things to check after merge:

1. A macOS/iOS release run's log does not report skipped signing or notarization. Those lines are the failure mode this PR exists to prevent, and they do not fail the job.
2. SSPlayerForGodot is on the repository access list of the org secrets. This repo lives under the SpriteStudio org, so the cri-middleware org secrets must be shared with it explicitly — the most likely thing to still be missing.