Chiapos

Added no bitfield plotting option (-e)

Added

- We now instrument temp space, time, and memory used building two k=27 on ci for performance tracking and comparison one for each back propagation method.

Changed

- We added a `-e` flag to allow you to choose between bitfield back propagation and the original back propagation. This brings performance a little ahead of the Beta 17 plotter release. However bitfield back propagation may perform faster on spinning disk. YMMV.
- More of the sorts in the bitfield sort use uniform sort for better performance.
- This version requires slightly more RAM at each k size.

Fixed

- Previous releases were not deleting temp files as they were no longer needed and therefor temp space needed for a k=32 went from 332 GiB to 499 GiB. This version only needs 332 GiB when using -u 64.
- Beta 17 style back prop had a buck that could occasionally lead to a segfault. We fixed that when using the `-e` flag.

Fixed

- We now close unused file-handles in the temp directory to avoid exceeding rlimit when using the new bitfield backprop and 128 sorting buckets.

Fixed

- Changed scope of match_indexes to make sure they are always reset even if R bucket is empty. This fixes "error 0" retrying in 5 minutes issues in certain edge cases.
- Windows plotting now correctly reports CPU utilization.

Added
- F1 generation is now fully parallel for a small speedup.
- We have bitfield optimized phase 2. There is only about a 1% increase in speed from this change but there is a 12% decrease in writes with a penalty of 3% more reads. More details in [PR 120](https://github.com/Chia-Network/chiapos/pull/120)
- You can now build chiapos with Cmake and MSVC. We know build and test C++ for Windows as well as create and test Windows wheels.

Changed
- We have moved from using gulrak/filesystem across all platforms to only using it on MacOS. It's required on MacOS as we are still targeting Mojave compatibility. This should resolve Windows path issues.
- Various pointers and memory management were modernized.

Fixed
- Starting with recent setuptools fixes, we can no longer pass an empty string to the linker on Windows when building binary wheels. Thanks jaraco for tracking this down.

Fixed
- When fixing a possible problem with reading past a buffer, we slowed down total performance by about 20%. This has an alternative fix that should maintain 0.12.31/Beta 15's plotting speed - if not speed up some based on other improvements in 0.12.32.
- Certain NTFS mount points could cause an error in gulrak/filesystem. We've implemented a patch and upstreamed the issue. See https://github.com/gulrak/filesystem/issues/75
- Added additional plotting testing to GitHub Actions
- Separated out the C++ build/test workflow from the Ubuntu wheel build and test.