Accept IPA input

[roedoejet]

Support aligning input that's already in IPA (e.g. eng-ipa)

Summary

readalongs align currently rejects IPA input. If you have a text file
that's already been g2p'ed into IPA (for example by a g2p engine that
isn't supported by g2p), there's no way to align it. We should support
-l eng-ipa (and other *-ipa langs) as valid alignment input.

Cause

get_arpabet_langs() in g2p/__init__.py explicitly filters out any
lang ending in -ipa, so these are never offered as valid inputs.

Proof of concept

Removing if not x.endswith("-ipa") in g2p/__init__.py makes readalongs align -l eng-ipa
work (confirmed working against a sandbox install of g2p)

Concerns for a proper solution

Simply un-filtering *-ipa langs pollutes readalongs langs, which
would then list every *-ipa variant (many with poor display names like
IPA, crg-ipa, crk-ipa). Ideally we'd accept *-ipa inputs
without listing them all in readalongs langs.

Priority

Low/medium

[joanise]

Additional consideration: allowing -ipa as the input means the users would see IPA rather than orthography. Would it not make more sense to have an alterative process that builds on readalongs make-xml, readalongs tokenize, and then overrides what readalongs g2p does?

Normally, the ARPABET is hidden from the user in an attribute, and I would expect to want to do the same thing for IPA.