Add optional AC-3 and E-AC-3 5.1 output

[BuffMcBigHuge]

Summary

• Adds opt-in AC-3 and E-AC-3 5.1 output modes while leaving PCM as the default.
• Probes the current Android audio route in automatic modes; force modes cover incomplete platform capability reporting.
• Encodes decoded 5.1 PCM on a bounded audio-priority worker and falls back one-way to the existing PCM renderer on initialization, route, encode, queue, or write failure.
• Vendors minimal LGPL FFmpeg n7.1.1 prebuilts for all four Artemis ABIs with a pinned, reproducible build script.
• Adds focused tests and runtime diagnostics without changing Sunshine, the GameStream protocol, or moonlight-common-c.

Motivation

Some Android TV and HDMI setups only deliver stereo or silence from multichannel PCM. Encoded Dolby output gives receivers another compatible 5.1 transport. Related to #22 and #283.

Testing

• :app:assembleNonRoot_gameDebug passed for armeabi-v7a, arm64-v8a, x86, and x86_64.
• :app:assembleNonRoot_gameRelease -x :app:lintVitalNonRoot_gameRelease passed for all four ABIs. The skipped gate contains 12 pre-existing values-ru extra-translation errors.
• Focused encoded-audio unit tests: 9/9 passed.
• Full unit suite: 55/60 passed. The five existing layout/theme, startup-context, and profile-navigation failures are unchanged from the branch baseline.
• Feature code has no Android lint errors. Repository-wide lint remains blocked by 26 pre-existing errors.
• git diff --check passed.
• The FFmpeg build was reproduced for every ABI and verified as LGPL-only with both ff_ac3_fixed_encoder and ff_eac3_encoder present.
• AC-3 transport was exercised successfully on a Sony A9G / HT-A3000 setup during development: 375/375 complete 1,792-byte access units were accepted.
• No Android target was connected during this final review, so no new on-device route/session test was run.

Notes

PCM remains the default, and every encoded option is marked experimental. AC-3 was stable in target testing. E-AC-3 produced 5.1 output, but the target setup previously showed brief intermittent dips. No dip-time trace was captured, so this PR does not include a speculative buffering change.