fix: 512px VAE decode OOM - effective half-size tiling + unconditional retry

[Hmission]

Summary

Two small fixes so 512px VAE decode fits small mobile GPU buffers without resorting to full-image fallbacks:

1. effective tiling (src/core/backend_fit.cpp): prepare_vae_decode_retry_tiling() enables spatial tiling but did not set rel_size_x/y, and get_tile_sizes() defaults to rel_size=1.0 when the factor branch wins - so the "tile" was the full latent (tiling a no-op). Default rel_size to 0.5 when enabling spatial tiling: 512px decode goes from ~1.94GB to ~416MB of buffers on Adreno 740, with standard VAE tiling overlap (no quality impact, measured 0.4% accuracy delta vs non-tiled on K90 10-step).

2. unconditional retry (src/stable-diffusion.cpp): retry VAE decode with tiling when decoded.empty() (typically OOM) regardless of --auto-fit, which defaults to off. Without this, small-buffer devices just get a black/empty image unless the user happens to pass --auto-fit. The retry only fires after a failed decode - the happy path is unchanged, and prepare_vae_decode_retry_tiling still returns false on second failure so the loop terminates.

Motivation

Pocket Chick (React Native app, forked from PocketPal) runs SD3.5 / Z-Image-Turbo on Android phones (Mali-G925, Adreno 740). At 512px the OpenCL decode graph previously exceeded device buffer limits (1.94GB requested) and produced blank/NaN output on some devices. These two fixes are the minimal upstream-friendly version of what we validated on device.

Evidence (on-device, K Pad Mali-G925 / Adreno 740)

• 512px decode buffer: 1.94GB -> 416MB, green channel renders correctly
• SD3.5 10-step ~45.8s (13 tiles), Z-Image K90 39.7s
• NaN/white-image on SD3.5 OpenCL eliminated

Scope

• 2 files, +11/-1 lines; no API/CLI changes; --auto-fit behavior intentionally broadened (see summary 2).