Qecp

- [x] publish `qecp` package to crate.io
- [x] add more advanced threshold evaluation script `benchmark/threshold_analyzer`
- [x] remove blossom V library in the repo to avoid license issue, now user has to download the blossom V package themselves
- [x] both UF decoder and MWPM decoder (new) support decoding erasure errors
- [x] start writing using `mdbook`

- [x] real-weighted union-find decoder
- [x] make union-find algorithm generic typed to enable advanced information to be computed efficiently
- [x] change blossom V library to use integer matching, because floating-point matching sometimes enter deadlock

- [x] build complete model graph using customized Dijkstra's algorithm
- [x] mwpm decoder
- [x] tailored mwpm decoder
- [x] remove all unnecessary .jobout, .joberror, sbatch files and clean up the git repo

- [x] remove `mwpm_approx.rs` because we don't need it anymore
- [x] redesign the simulator for improved locality and faster simulation (the current bottleneck), use Box for optional fields, for example correlated error rate

- [x] remove `BatchZxError`, `ZxError`, `ZxCorrection`, `ZxMeasurement` and several obsolete types and related functions
- [x] delete unused and obsolete benchmark tools
- [x] remove features `MWPM_reverse_order` and `MWPM_shuffle`
- [x] remove feature `reordered_css_gates` because it's now the default option
- [x] remove feature `noserver` because it doesn't work as expected to speed up compilation

- [x] design a set of profiling test cases to both test correctness of existing decoders and record performance for future optimization compares
- [x] upgrade dependencies to latest version