Ingest all sentence alignments

[jacobwegner]

A downside of "tokens or bust" (and alluded to in the "continuum" discussion in scaife-viewer/backend/#19) is that we can't display any type of an alignment unless we've mapped it down to the token level.

We could feasibly reintroduce the "blob" approach for GRC's sentence alignment or certain Hafez alignments. We'd still want to map versions and text parts, but we would end up just using the first "token" of a text part and render a simple blob.

The "renderer" hint to the front end could instruct it to group by records and use the record's HTML.

This would allow us to display the sentence alignments as-is without doing a ton of additional parsing.

[jacobwegner]

At any rate, I might try a parsing pass of 1.1-1.25 from the sentence alignments to see how feasible that ends up being

[jacobwegner]

(Might actually defer this until we can "fully" load syntax trees to the token level, as the GRC alignment is tied to syntax tree serial numbers and might be able to otherwise to do some regexes to map to tokens)