feat(sampler): Polyphonic voice allocation, live internal resampling, and multi-velocity layers

[delaudio]

Overview & Motivation

The current trk-sampler is primarily focused on monophonic sample triggers per track. Expanding the sampler engine with polyphonic voice stealing, internal live resampling, and multi-velocity keyzones makes trk capable of producing lush acoustic pianos, pads, chords, and SP-404 style sound design.

Proposed Features

1. Polyphonic Voice Allocation (4 to 16 voices per track):
   • Configurable polyphony modes: Mono, Legato, Poly 4, Poly 8, Poly 16.
   • Intelligent voice stealing strategies: Oldest, Lowest Volume, Same Note Retrigger.
2. Internal Live Resampling (SP-404 / Elektron Style):
   • Ability to record pattern or master audio output internally directly into an empty sample slot.
   • 1-key bounce: resample active pattern with all active DSP effects into a new WAV loop.
3. Multi-Velocity Sample Layering:
   • Assign multiple WAV files across velocity zones (e.g. Vel 0-63 soft hit, Vel 64-127 hard hit).
   • Support for basic SFZ multi-sample mapping descriptors.

Acceptance Criteria

• Sampler tracks can trigger and sustain multiple overlapping voices simultaneously.
• Voice stealing gracefully avoids audio clicks using sub-millisecond fade-outs.
• Resampling captures internal audio cleanly into an allocated sample slot.
• Velocity layering selects appropriate sample based on note velocity.
• Audio unit tests covering voice allocation and resampling buffer accuracy.