Add Android input preset selection

[Avejack]

Closes #44

Summary

This adds a small Android-specific capture input preset option to Recorder.init:

await Recorder.instance.init(
  androidInputPreset: AndroidInputPreset.voiceCommunication,
);

The default remains unchanged: if androidInputPreset is omitted/null, miniaudio leaves the Android input preset unset and the plugin keeps the previous platform-default behavior.

Changes

• Add public AndroidInputPreset enum with the common AAudio/OpenSL-backed presets:
  • generic
  • camcorder
  • voiceRecognition
  • voiceCommunication
  • unprocessed
• Extend the Dart init path and native FFI init ABI with an Android preset integer.
• Map the value on Android to both:
  • deviceConfig.aaudio.inputPreset
  • deviceConfig.opensl.recordingPreset
• Keep non-Android platforms, including web, as documented no-ops.
• Add a small Android-only selector in the example app for manual comparison.
• Update README/API docs and add an Unreleased changelog entry.
• Add focused tests for stable enum values and null/default mapping.

Validation

Automated checks run locally:

• dart analyze -> no issues found
• flutter analyze --no-pub example -> no issues found
• flutter test --no-pub test/android_input_preset_test.dart -> all tests passed
• cmake -S src -B /tmp/flutter_recorder_build && cmake --build /tmp/flutter_recorder_build -> native target built successfully
• git diff --check -> clean

Manual Android testing:

• Tested on Lenovo TB328FU.
• Compared the current default path against voiceRecognition, voiceCommunication, camcorder, and unprocessed.
• voiceCommunication produced the needed louder/fuller speech capture on this device, while the default remains unchanged for existing users.

Scope

This intentionally does not add a larger audio_session integration. audio_session remains useful for session/focus/route management, but Android input preset selection needs to be passed to miniaudio's capture config before device initialization.

[Avejack]

Added a follow-up stabilization commit after device testing on Lenovo TB328FU.

The original AAudio voiceCommunication path could intermittently fail ma_device_start() even though Android moved the stream to STARTED immediately afterward. The follow-up commit routes the Android default-device voiceCommunication preset through OpenSL, which supports the same VOICE_COMMUNICATION recording preset and avoids the observed AAudio state-transition race on this OEM path.

Additional validation after the follow-up commit:

• dart analyze -> no issues found
• flutter analyze --no-pub example -> no issues found
• flutter test --no-pub test/android_input_preset_test.dart -> all tests passed
• cmake -S src -B /tmp/flutter_recorder_build && cmake --build /tmp/flutter_recorder_build -> native target built successfully
• Manual Android repeat test on Lenovo TB328FU: voiceCommunication with repeated init -> start -> deinit now starts reliably.

[alnitak]

Thanks a lot @Avejack! This PR is much appreciated.

I didn't know about mic presets. I tried this and it works great. I found that on my Samsung Galaxy S22, using voiceRecognition and unprocessed, the mic volume is louder, but this depends, I think, on the brand manufacturer.

I should give this repo more consideration. If you find any other features or fixes, please feel free to make new PRs :).

LGTM, merging and releasing an update.