storage: improve documentation of fields in MVCCStats

MVCCStats fields are confusing. The relationship that committed MVCC versions, provisional MVCC versions, range keys, intents, and replicated locks have on the different fields is difficult to follow. We should improve the code commentary of the struct, ideally with examples and diagrams to demonstrate the different inputs to the statistic values.

From @sumeerbhola:
> I think part of the problem is that the code comments in MVCCStats are written as if (a) the scheme is normal and obvious, (b) there is no example showing older versions, latest version, latest provisional value with an intent and clearly saying where the various bytes will be accounted.

Jira issue: CRDB-31883