Docs: Extend PruningPredicate with background and implementation info

[alamb]

Draft as it builds on #9183

Which issue does this PR close?

related to #9171

Rationale for this change

We use the PruningPredicate logic extensively in InfluxDB and we have found it is not an easy concept to understand. I have struggled to explain it face to face concisely so I wrote it down with the help of @appletreeisyellow

I would like to add this content to help both our future engineers as well as the broader community understand how this code works (perhaps so they can extend it)

I also harbor an ulterior motive that anyone who works on / with DataFusion can get an accessable master class in practical database implementation techniques by reading our docs

What changes are included in this PR?

Add background and algorithm description to PruningPredicate docs

Are these changes tested?

By doc test

Are there any user-facing changes?

More docs. No code changes

[appletreeisyellow]

🤣

[appletreeisyellow]

Rows that evaluate to NULL are NOT returned (also “filtered out”)

Seems like this is the opposite of what discussed below.

For example, the quote below says -- "containers that evaluate to NULL are returned (also 'kept')"

/// * NULL: there MAY be rows that pass the predicate, KEEPS the container
/// Note that rewritten predicate can evaluate to NULL when some of
/// the min/max values are not known.

Did I misunderstand something here? 🤔

[alamb]

This is a very good observation.

Yes the NULL semantics in filter expressions is the opposite of what the rewritten pruning predicate does with NULL (which is quite confusing). I added a note trying to clarify this: 74261e8

[Jefffrey]

Makes sense to me, thanks for the great explanation. I'm not familiar with the PruningPredicate code much, but from the documentation alone here it looks good (reviewing from phone so it's a little hard to check out the code 😅)