Support integer and mono audio output formats

[brxs]

Part of #107.

Adds cross-platform output-device compatibility at the CPAL boundary while preserving the engine's internal 48 kHz stereo f32 contract.

What changed

• supports f32, i16, and u16 output streams
• handles mono downmix, stereo output, and wider master/cue routing
• clips and converts samples without allocating in the realtime callback
• processes arbitrarily large host callbacks in bounded preallocated tiles, including synchronized master/cue and resampled output
• keeps non-48 kHz resampling ahead of final sample conversion
• reports the selected output sample format
• uses cross-toolchain MXCSR access for FTZ/DAZ on x86_64

Verification

• 16 focused device tests, including oversized primary, master/cue, and resampled callbacks
• full engine suite: 116 tests
• cargo clippy --all-targets -- -D warnings
• Windows MSVC engine compile check
• full Rust workspace compile check

Actual playback qualification on Windows/Linux mono and integer-only devices remains part of the platform hardware matrix; this PR establishes and unit-tests the portable format/channel contract.

[brxs]

Superseded by and fully consolidated into #131. The combined branch includes this lane's substantive changes, subsequent integration fixes, and exact-head cross-platform validation. Please continue review in #131.