Immunedb

* Minor fix for V-identity when exporting clones.

* Fix for clonal similarity.
* The Docker container now proxies both the frontend and API through port 8080.
Instead of running `serve_immunedb.sh`, simply navigate to
http://localhost:8080/frontend/`db_name`.
* Exporting clones now includes a `v_identity` column.

* Identity filtering for all CLI commands has been changed to use fractions
rather than integers.
* The `similarity` clonal assignment method now has a `--level` flag taking
either `aa` (the default) or `nt` which sets the level at which similarity is
calculated.
* Due to the change above, the `tcell` clonal assignment method has been
removed. Use `similarity` with the flags `--level nt --min-similarity 1`
for the same results.

* J-genes are now assigned based on percent identity rather than number of
matching bases which favored longer J-genes.
* Exporting has been cleaned up and the API and CLI have been unified.
* Clones can now be pooled when exported.
* Exporting via the API now returns a `uid` token rather than the result. The
result can be accessed once the data is ready using the `uid`.
* A new `overlap` export format has been added both to the CLI and
web-interface.

* Exporting has been entirely rewritten. Exporting from the CLI and
web-interface via the API now use unified methods.
* Duplicate sequences at the sample level are no longer stored. This increase
identification / importing speed drastically.
* Selection pressure, sample metadata, clones summaries, and clonal overlap
have all been added to `immunedb_export`.
* `immunedb_clone_trees` now has new parameters to filter mutations & sequences
by copy number and number of samples.
* Importing for Adaptive data has now been added to the `immunedb_import`
command.
* Metadata can now be updated and samples can be combined with the
`immunedb_modify` command.

* ImmuneDB has now been ported to Python 3. ImmuneDB v0.23.0 will be
maintained in a separate branch for bug fixes only. No new Python 2 features
will be added.
* Multiprocessing for identification has been rewritten to process each sample
sequentially.
* Fixed a bug where clone counts were incorrect during API call.
* Adding option to export in AIRR format.
* Clearcut must now be located in $PATH.
* The Docker container has been rewritten and no longer uses Docker Compose.