Add option for checking w3c specification on expand()

[cthoyt]

References #103.

This PR adds a w3c_validation flag to instantiation of a converter and to Converter.expand() that turns on W3C CURIE specification validation. For now, it's not a "true" implementation of the W3C standard since this will take a huge amount of effort to implement and test (for, IMO, small returns), but it hits the main pain points Chris mentioned in #103 (specifically, the square bracket issue).

This PR also begins developing a benchmarking suite of CURIEs (both valid and invalid)

Demo

This can be used when expanding a CURIE:

import curies

converter = curies.Converter.from_prefix_map({
    "smiles": "https://bioregistry.io/smiles:",
})

>>> converter.expand("smiles:CC(=O)NC([H])(C)C(=O)O")
https://bioregistry.io/smiles:CC(=O)NC([H])(C)C(=O)O

>>> converter.expand("smiles:CC(=O)NC([H])(C)C(=O)O", w3c_validation=True)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/cthoyt/dev/curies/src/curies/api.py", line 1362, in expand
    raise W3CValidationError(f"CURIE is not valid under W3C spec: {curie}")
W3CValidationError: CURIE is not valid under W3C spec: smiles:CC(=O)NC([H])(C)C(=O)O

This can be used when instantiating a converter:

import curies

>>> curies.Converter.from_prefix_map(
...     {"4dn.biosource": "https://data.4dnucleome.org/biosources/"},
...     w3c_validation=True,
... )
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/cthoyt/dev/curies/src/curies/api.py", line 816, in from_prefix_map
    return cls(
           ^^^^
  File "/Users/cthoyt/dev/curies/src/curies/api.py", line 527, in __init__
    raise W3CValidationError(f"Records not conforming to W3C:\n\n{msg}")
curies.api.W3CValidationError: Records not conforming to W3C:

  - Record(prefix='4dn.biosource', uri_prefix='https://data.4dnucleome.org/biosources/', prefix_synonyms=[], uri_prefix_synonyms=[], pattern=None)

Caveats

• there are several other places where this flag could propagate if it seems like a good idea in the end
• the actual conformance checking could be improved to reuse some of what's in LinkML, but it needs to be made more readable, understandable, and thoroughly tested. https://github.com/dgerber/rfc3987/blob/gh-archived/rfc3987.py is even more detailed