Actipy

What's Changed
- Fix bug in `detect_nonwear`: When input file is too small, `stationary_len` can become an empty series whose `dtype` defaults to float in Pandas, making comparison with timedelta types fail.

What's Changed
- Fix bug: force copy when loading parsed data, otherwise cleanup of temporary files fails 6a32d7ebcec90d1f6c1f506794c3b2ab1813afb2

What's Changed
- Reverted major changes since `v0.0.2` involving out-of-memory computation using `memmap` and `tempfile` because these were causing issues in Windows, and sometimes in Linux-based clusters. This is a temporary resolution while we look into this.

What's Changed
- **Minimal required version is now Python 3.8**. This is to be able to require `numpy>=1.22` which has a vulnerability patch. 3f8ae33478025ee9c37c84bf4a06d910f94342fb
- More flexible dependencies 3f8ae33478025ee9c37c84bf4a06d910f94342fb
- Include `extra_requires=...` in setup.py. Now can use `pip install actipy[dev]`, `pip install actipy[dev,docs]` etc. for easier dev setup. 7e47fffedae9b76ce8f424435dd90c9f2da234ca

This version is BROKEN due to a wrong compiled java code. Use 1.0 or 1.1 instead.

What's Changed
* Fix critical bug when not enough calibration samples crashed program.
* Minor improvements to docs, fix typos, etc.

Fixes
- Fixed calibration bug when a series of constant values confuses statsmodels fb8164aaff9d015d18212b996f68cc47400cb671

Improvements
- Improved performance of `npy2df` 7c6a3b6b9d2e03c89d7aabd059bdb76b4f0790ed
- `lowpass` now doesn't revert the uniform resampling after filtering as it's too costly 953917c46bb724ee7e97f874fd16cde8443d8a89
- Skip resampling if the rate is already satisfied
- Better resampling, now properly supporting non-integer rates 66eb1fe993ecfec45631ff08ee6d535c24649b55
- Many other major works to improve memory consumption of all routines, including using chunking and memmapping.

Other changes
- Deprecate `misc` metrics 88e60d9a959e3bb26944dfd742fb6a614c3ecba9
- Rename temperature column name `T` -> `temperature`