feat: smart shuffle recommendation engine

[MunishMummadi]

Note: This is full AI generated PR. But I understand majority of the code. Since I am not good at kotlin at all. I am using AI. You have everyright to decline to review or close this PR. But I would love to answer any questions and do as many changes as I can to get this in.

What is it?

• Bugfix (user facing)
• Feature (user facing)
• Codebase improvement (dev facing)
• Meta improvement to the project (dev facing)

Description of changes

Got the idea from xikipedia
Adds an offline Smart Shuffle recommendation engine for local libraries (phone + Android Auto).

• Learns from real playback signals (early skip, mid abandon, solid listen, near-complete, replay) plus an explicit Like
• Builds shuffle queues from album/artist/genre/decade/duration feature weights (### not a fake RNG “For You” screen)
• Marks repeatedly skipped songs as undesirable and surfaces them under Settings → Songs to review (forgive or delete)
• Integrates into existing shuffle-all playback so Android Auto still reports shuffle ON while ExoPlayer uses an ordered smart queue
• Persists the model locally (smart_shuffle_model.json) with debounced/atomic saves
• Mobile: Like button on the playback panel
• Android Auto: MediaSession custom “Like” action (outline/filled heart)

Fixes the following issues

Any additional information

APK testing

app-debug.apk.zip

Due diligence

• I have read the Contribution Guidelines.
• I have read the Why Are These Features Missing? page.