Semibin

* Fix issue 93 more thoroughly (see 101)

* Fix edge case when calling prodigal with more threads than contigs (93)

* More balanced file split when calling prodigal in parallel
* Fix bug when long stretches of Ns are present (87)
* Better error messages (90 & 91)

* Improve `check_install` command by printing out paths and correctly
handling optionality of FragGeneScan/prodigal
* Reuse markers.hmmout to make the training from several samples faster
* Add option `--tmpdir` to set temporary directory
* Substitute FragGeneScan with Prodigal (FragGeneScan can still be used
with `--orf-finder` parameter)
* Add 'concatenate_fasta' command to combine fasta files for multi-sample binning

* Provide pretrained models from soil, cat gut, human oral,
pig gut, mouse gut, built environment, wastewater and global (training
from all environments).
* Add `check_install` command and run `check_install` before easy* command
* The user can now specify a pre-computed contig annotation table in
mmseqs format
* Fix bug with non-standard characters in sample names (68)
* Better subcommand names (`generate_sequence_features_*` and
`generate_cannot_links`)

* Faster `SemiBin --version`
* Lower memory usage and faster speed for `bin` subcommand
* Reclustering is now the default (due to improved speed). Added
`--no-recluster` option to disable it
* Output of bedtools is now processed as a stream instead of using a
(potentially large) intermediate file
* Fix bug with --min-len. Previously, only contigs greater than the given
minimal length were used (instead of greater-equal to the minimal length).
* Fix bugs downloading GTDB
* Respect $XDG_CACHE_DIR if set
* Implement CACHEDIR.TAG protocol for the SemiBin cache directory