v0.3.27

## What's Changed
* fix(confidence): clear stale low_confidence flag on re-evaluation (#910) by @Sarthak-Pandey in https://github.com/NandhaKishorM/laya/pull/911
* fix(router): accept and compose per-call hooks in predict_batch and route_batch (#909) by @Swatantra-66 in https://github.com/NandhaKishorM/laya/pull/912
* fix(email): match French device footers with normal spacing by @AzarudeenshariffA in https://github.com/NandhaKishorM/laya/pull/913
* fix(router): run predict_long's scan after the hook chain, not as a hook by @keemsisi in https://github.com/NandhaKishorM/laya/pull/914
* fix(evals): a coverage cut must not split a group of tied confidences by @keemsisi in https://github.com/NandhaKishorM/laya/pull/918
* fix(calibrate): fit abstention thresholds on the scale the gate will read by @AlKor13 in https://github.com/NandhaKishorM/laya/pull/920
* fix: keep transformers from importing TensorFlow, which crashes a load when it cannot load by @PerryLink in https://github.com/NandhaKishorM/laya/pull/924
* feat(java): laya-java: a JVM inference package, gated against the Python reference by @keemsisi in https://github.com/NandhaKishorM/laya/pull/927
* docs(benchmarks): MPS fp16 autocast against fp32 on an Apple M1 Pro by @cacheline999 in https://github.com/NandhaKishorM/laya/pull/928

## New Contributors
* @AzarudeenshariffA made their first contribution in https://github.com/NandhaKishorM/laya/pull/913
* @cacheline999 made their first contribution in https://github.com/NandhaKishorM/laya/pull/928

**Full Changelog**: https://github.com/NandhaKishorM/laya/compare/v0.3.26...v0.3.27