fix(android): preserve manually set releaseIdentifier

[cat-ph]

💡 Motivation and Context

PostHogMetaPropertiesApplier.applyToConfig had an inverted guard: the comment said "if releaseIdentifier is already set, we don't need to do anything", but the condition overwrote a manually configured releaseIdentifier with the auto-generated fallback (applicationId@versionName+versionCode). Anyone setting releaseIdentifier in PostHogAndroidConfig (as documented) lost it on init, breaking proguard symbolication due to the map-id mismatch.

This fixes the precedence to:

1. A pre-set releaseIdentifier is preserved (the asset read is skipped entirely).
2. Otherwise the map-id from posthog-meta.properties is used when present.
3. Otherwise the fallback is applied — including the previously broken case where the meta file exists but contains no map-id, which used to leave releaseIdentifier null.

💚 How did you test it?

• New unit tests in PostHogMetaPropertiesApplierTest covering all five paths (preserve manual value, use meta map-id, fallback on missing file / missing key / read error).
• Full unit test gate (./gradlew testDebugUnitTest) and spotlessCheck pass.

📝 Checklist

• I reviewed the submitted code.
• I added tests to verify the changes.
• I updated the docs if needed.
• No breaking change or entry added to the changelog.

If releasing new changes

• Ran pnpm changeset to generate a changeset file

[cat-ph]

this is actually a more important case I saw in prod, tens of thousands of frames per week here at a first/quick look (although some could be older versions who don't have a releaseIdentifier in the first place).

the first one with manual releaseIdentifier isn't as common

[greptile-apps]

_{Reviews (1): Last reviewed commit: "fix(android): preserve manually set rele..." | Re-trigger Greptile}

[marandaneto]

will check, i remember testing this so quite surprise its a bug, maybe a regression

[marandaneto]

Was it a regression?

Your memory is correct:

• The initial implementation in PR chore: proguard support #316, commit 00d4e33 from November 5, correctly did:

    if (config.releaseIdentifier.isNullOrEmpty()) {
        config.releaseIdentifier = fallback
    }

• The November 28 refactor in commit bd1373f introduced PostHogMetaPropertiesApplier and accidentally inverted that
  behavior.
• It shipped in posthog-android 3.27.0.

So it was a regression during development of PR #316, likely after you tested it. Strictly speaking, it was broken in
the first stable release containing releaseIdentifier, so it was not a regression between stable SDK releases.