feat(web): add new specifiable model method for validating contexts for autocorrect

[jahorton]

As noted by #12648, users will not expect autocorrect to trigger for all contexts. For example, in English, users generally don't expect autocorrect to trigger when typing numbers. That said, this can be language-dependent - there are some languages that use digits as letters in some contexts. (#12877 (comment))

To facilitate good all-around handling, #12877 implemented a default behavior similar to that of programming identifiers as a baseline for most languages. For cases like those alluded to with the linked comment, we also wish to add space for individual lexical models to include a method that overrides this behavior. The method would then return a boolean value indicating whether or not autocorrect is considered valid for the current token.