schema: Specify the encoding for character offsets

[varungandhi-src]

This PR adds a new PositionEncoding field so that indexers can specify
what type of character offsets they are using. This way, consumers of SCIP
can unambiguously interpret the offsets for non-ASCII data.

I have kept it as a field on Document rather than Index because:

1. There is no additional benefit from having it on Index because
   occurrences only belong inside Documents, not outside.
2. It allows one to concatenate indexes from different sources
   which use different kinds of offsets.

Test plan

n/a

[olafurpg]

Good point about making this per document. Would make sense to support the same for the text document encoding.