Reduce the memory allocation during search with mutable IndexedScore.

[isoos]

• Follow-up to Improve token index performance by using fewer memory during search. #8152.
• Fewer operation is required as document weight is part of the pre-calculation at index creation. Some scores in test did change, I think these changes are because of the changes of the generic method that is being used in search and they are fine.
• This change does not completely eliminate the Map-based immutable Score, but reduces its use (and its map-building operations) by extending the use of the list-based score across multiple token index and multiple words (fewer list allocations).
• Local search benchmark shows about 3-6% total query time improvements.
• Note: I'm also planning a follow-up PR that expands the concept on the filtering that happens before the text search part, possibly getting further benefits.

[isoos]

Note: after some offline discussion, I'll split this PR into multiple parts, leaving this as a draft for now.

[isoos]

Closing, as all parts have been submitted for review or merged already.