Ctakesclient

This release adds the ability to query the `termexists` cNLP transformer when listing polarity. It's another polarity model that's similar to the default `negation` model, but may be a bit better.

The model can be installed from [docker hub](https://hub.docker.com/r/smartonfhir/cnlp-transformers/tags?page=1&name=termexists).

Set `URL_CNLP_TERM_EXISTS` to point at the correct endpoint (defaults to `http://localhost:8000/termexists/process`).

Use it like so:

from ctakesclient.transformer import TransformerModel, list_polarity

list_polarity(text, spans, model=TransformerModel.TERM_EXISTS)

Whoops, one more tweak to the ICD10 codes, with another major version bump with it.

No Python changes, but the ICD10 list for covid symptoms has been modified. So just be aware of that before you upgrade, as it may affect clinical results based on it.

A few breaking changes in this major release:
- Support for Python 3.7 has been dropped (slightly before it goes end-of-life)
- All network calls have been changed to be async methods. Specifically, the following API calls are now async:
- client.post
- client.extract
- transformer.list_polarity
- transformer.map_polarity

These new async calls also take an optional `client` kwarg parameter, to provide an existing
`httpx.AsyncClient` session.

Fixes accidentally duplicating CUI codes in text2fhir output.

Add `filesystem.covid_symptoms_path()` and `filesystem.umls_semantic_groups_path()` to get direct access to the `.bsv` files for each.