Implement background pipeline cache storage

docs/quest-port.md

@@ -603,6 +603,7 @@ Bring-up fixes that only a device could reveal:
 | Crypto++ `cpu-features.h` not found | The NDK ships cpu-features as source | Compiled into `mkw_cryptopp` on Android |
 | Exploded racers and menu characters; smeared movie panels in the menus; then, once those were fixed, damaged eyes and slightly misplaced detail on characters | The Adreno 740 driver reads the wrong bytes when the shader multiplies an index by a stride that is not a multiple of 4. That covers the vertex fetch (`ubuf.vtx_start + vidx * stride + offset`) and indexed array reads (`array_start + index * stride`, e.g. 6-byte S16 normals). GX packs both byte-tight, so skinned models (a 1-byte `PNMTXIDX` first, stride 7) broke everywhere | Android pads every uploaded vertex and every indexed-array element to a 4-byte stride (`padded_upload_stride` in `lib/gx/gx.cpp`). Offsets inside a vertex or element are unchanged, and desktop is unchanged. **Fixed, headset-verified 2026-09-16** at character select and a Grand Prix start |
 | Every launch recompiled every shader: a 14 to 34 s prewarm, and the Dawn blob cache reporting exactly one miss and no stores | Dawn's monolithic Vulkan pipeline cache is only written by `PerformIdleTasks`, which `gpu.cpp` resolved through the Windows DLL alone, and the quit path ends the process without `aurora_shutdown`, so nothing compiled after prewarm was kept either | Static Dawn calls it directly; `aurora_store_pipeline_caches` runs at a race exit, when the session loses focus and on the quit path, and the compiler stores idle bursts itself while the headset shows the virtual screen. The unpacked `initial_pipeline_cache.db` is also refreshed per APK install now |
+| The game and its music froze for 0.9 to 3.8 s at the end of a race, longer as a session went on | Dawn serializes its monolithic Vulkan pipeline cache (58 MiB on the Quest, and it barely compresses) under its device lock and hands it to Aurora's blob-cache callback, which compressed it and committed it to SQLite still inside that lock; the frame worker waited on the device and the game thread on the frame worker. The store also ran on the XR pacing thread | `gpu_cache.cpp` copies the blob and queues it; a writer thread compresses and commits, and bytes identical to what the database holds are not rewritten. The race exit uses `aurora_request_pipeline_cache_store`, off the pacing thread. The log's `Stored the pipeline caches` line reports the part still spent holding the device |
 
 How the explosion was isolated, so the next Adreno rendering bug starts further
 ahead: