vParquet5: remove well-known attribute columns

[stoewer]

What this PR does:

This PR consists of the following changes:

• Remove the existing columns for well-known attribute columns
• Apply a default dedicated attribute column config when no dedicated columns are present
• Encodings vp4 and vp3 ignore dedicated column configs that match existing well-known columns. This helps to avoid conflicts with the new default dedicated columns in older encodings
• Use LRU cache for unmarshaled dedicated columns and avoid allocations for cache lookup: This is a small optimization to compensate for the larger and more frequently used dedicated column configs. I plan to implement further optimizations in a separate PR, if this becomes a problem.

Which issue(s) this PR fixes:
Contributes to #5665

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[mdisibio]

Looking really good, I think this migration path for vp4 to vp5 without performance penalty or interruption is great. Left a few qs and comments.