fix(android): migrate CargoKit to Android Components

[HsukqiLee]

What changed

• migrate CargoKit from legacy applicationVariants/libraryVariants to AndroidComponentsExtension.onVariants
• parse the numeric component of compileSdkVersion so AGP values such as android-37.0 remain supported

Validation

• verified fjs CargoKit debug/profile/release tasks under AGP 9.4.0-rc01 and Gradle 9.6.0
• verified app configuration completes without the AGP applicationVariants/newDsl warning after suppression settings are applied

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[iota9star]

Hi @HsukqiLee — thanks for the PR. A note on the CI you're seeing here: that run never actually started any jobs. Since the secret passthrough landed on main (Aug 13), every Build All Platforms run has failed at startup because precompile-binaries.yml didn't declare PRECOMPILE_BINARIES_PRIVATE_KEY under workflow_call while the caller passes it explicitly — GitHub rejects the workflow with "Invalid secret, PRECOMPILE_BINARIES_PRIVATE_KEY is not defined in the referenced workflow." Your run then sat queued for 30 days until GitHub force-failed it. It's unrelated to your changes.

The startup fix is up in #25 (verified by dispatching the workflow). Once that merges, please rebase this PR so CI can actually validate the Android Components migration. The migration itself looks right to me on read-through — onVariants + numeric compileSdkVersion parsing keeps both the AGP 8.7.3 and 9.0.1 compatibility legs in scope. The main thing CI needs to prove is that the no-cargokit-warnings gate still holds on both legs and that jniLibs wiring from inside onVariants is picked up.

[HsukqiLee]

Thanks for the detailed rundown and for merging main into the branch, @iota9star — the workflow now actually starts (prepare is green and the quality gates are running). I'll keep an eye on the no-cargokit-warnings gate across both AGP legs and the jniLibs wiring from inside onVariants.

Also, with #25/#26/#27 now on main, the changes we were carrying in our fork for the secret declaration, the h2/rustls security bumps, the flaky close-wins-while-initializing test, and the FRB 2.13 upgrade are all superseded upstream — nothing further needed from us there.

One more independent fix carried in our fork, split out as its own PR: #29 (keeps precompiled snapshot staging paths under the Windows MAX_PATH limit).