Methylprep

- Fixed bug in CSV export; quality_mask (probes to be excluded from pickles) were NaN instead of 1.0

- Merely a maintenance release to deal with some dependency conflicts in the python science stack.
Pandas version 1.3+ now supported.

- completely rewritten and updated documentation, including a more extensive tutorial
- updated all manifests to include probe-to-locus mapping for two genome builds
- Uses OLD_ in front of 4 genomic columns to denote the older genome build in each respective array/organism
- mouse manifest has mm39 (newer) and mm10 (older) genome assemblies.
- human hg38 is a corrected and improved version of (OLD) hg19.
- 27k array is no longer supported. (It turns out it was never supported, but we just never unit-tested it. And
because there are no control probes and no type-II probes on the 27k first generation array, it would be a lot
of work to support it, and nobody has ever asked about it.)
- removed read_geo and detect_header_pattern from methylprep, because it is maintained in methylcheck and imported
into methylprep now.
- new beta_bake CLI option `-z / --compress` will put all downloaded files into a zipfile. This used to be the default
behavior, and now it is optional.
- fixed minor bug where malformed custom user-supplied manifest files will return a useful error message.
- better processing error detection with a `check_for_probe_loss` functions that warns if probes are dropped

- Fixed Reading IDATs progress bar in 'process'
- `download` now uses HTTPS requests to fetch GSExxxxxx_RAW.TAR data files instead of FTP, because this is way more reliable and avoids errors.
- Better error messages when trying to download and process GEO datasets containing multiple array types. The process splits these samples into separate folders based on the meta data, and tries to run each array type separately.
- Improved documentation everywhere
- Improved support for GEO series_matrix.txt.gz files and _family.xml -tbl-1.txt files
- Fixed bug where quality_mask was removing SNP (rs) probes from CSV or SampleDataContainer output.
- Removed detect_header_pattern and read_geo from methylprep. These are maintained in methylcheck and
not called anywhere. Instead, methylprep tries to import methylcheck in the Download functions that need
to read GEO data files. But if a user has not installed methylcheck, they will only see an error if they
run specific commands in Download that require it.

- Bug fix: added 'Extended_Type' into control_probes.pkl output. Required by methylcheck.plot_controls().
- Minor bug fixes and improved unit test coverage.
- Fixed bug where `process --minfi` was not working with `--all`. Added more test coverage for CLI.
- updated read_geo to handle more edge cases
- deprecated some never-used functions.
- instead of methylprep.files.idat.RunInfo use IdatDataset( verbose=True )

- MAJOR refactor/overhaul of all the internal classes. This was necessary to fully support the mouse array.
- new SigSet class object that mirror's sesame's SigSet and SigDF object.
- Combines idats, manifest, and sample sheet into one object that is inherited by SampleDataContainer
- RawDataset, MethylationDataset, ProbeSubtype all deprecated and replaced by SigSet
- SampleDataContainer class is now basically the SigSet plus all pipeline processing settings
- new mouse manifest covers all probes and matches sesame's output
- Processing will work even if a batch of IDATs have differing probe counts for same array_type, though those
differing probes in question may not be saved.
- unit tests confirm that methylprep, sesame, and minfi beta values output match to within 1% of each other now. Note that the intermediate stages of processing (after NOOB and after DYE) do not match
with sesame in this version. Can be +/- 100 intensity units, likely due to differences in order of
steps and/or oob/mask probes used.