Openst

Various fixes in flowcell map and image stitching.

Version used in STAR Protocol.

What's Changed
* Add mln example by danilexn in https://github.com/rajewsky-lab/openst/pull/71


**Full Changelog**: https://github.com/rajewsky-lab/openst/compare/v0.2.2...v0.2.3

In `flowcell_map`, writes the `lanes_and_tiles.txt` file to the output directory, in case the bcl folder does not have write permission.

Bugfix when using many cores for `flowcell_map`

Added
- New `flowcell_map` command to convert basecalls from a whole flow cell into barcodes-to-spatial coordinates maps
- Support for `bcl-convert` in addition to `bcl2fastq` for demultiplexing
- Added support for the Axio microscope in `image_stitch` command
- Implemented a more efficient tiling and stitching process for image preprocessing
- Added `dnaio` dependency for faster FASTQ parsing

Changed
- Updated `barcode_preprocessing` to handle larger datasets more efficiently
- Improved memory management in various parts of the pipeline
- Refactored `pairwise_aligner` for better performance and reliability
- Updated segment to use OpenST-specific model directory
- Modified `transcript_assign` to handle edge cases better
- Improved error handling and logging across the package
- Updated Python version requirement to >3.8 and <3.12

Fixed & Improved
- Corrected issues with coordinate systems in several functions
- Enhanced documentation
- Optimized performance for large dataset processing

What's changed
- `pairwise_aligner`: cleanup of cli, does not perform rotation/flipping on image but on ST coordinates (avoid redundant resaving of image). Consider tile bounds when staining image and ST coordinates differ.
- `manual_pairwise_aligner`: fix rendering depending on threshold, proper flipping for fine alignment. Assign coarse if tile falls outside of image.
- `apply_transform`: do not perform flipping when supported by < 3 keypoints.


**Full Changelog**: https://github.com/rajewsky-lab/openst/compare/v0.0.11...v0.1.2

This release fixes the implementation of `openst image_preprocess`, and makes models available in bimbstatic (as for our `HE_cellpose_rajewsky`).