Reduce internal code to array-only concepts

[kbrock]

Description

Prerequisite for the serializer: internal methods now work with
ancestor_ids (array) instead of the raw column string. Only the
parse/generate boundary knows about the format.

Before

Builder methods accessed the raw column directly and parsed inline.
Each method carried format knowledge (root literal, delimiter parsing).
Array format went through the same parse/generate path as strings.

After

Internal methods delegate to ancestor_ids boundary methods.
child_ancestry_value removed from format modules (only caller was
child_ancestry, which now uses generate_ancestry(path_ids)).
Array format with ancestor_ids column skips parse/generate entirely.

What this does NOT do

• No serializer (Rails Type) yet
• No validator changes
• No public API changes
• No changes to SQL generation or scopes

Non-obvious changes for developers

• Array format now expects ancestor_ids as the column name.
• child_ancestry_value is gone from format modules. Custom format
  subclasses should implement generate instead.
• @_ancestor_ids ivar cache bypassed for array format. Goes away
  when serializer lands.

Type of Change

• Feature

Checklist

• My code follows the style guidelines
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally

How Has This Been Tested

• bundle exec rake test (mp1 sqlite)
• FORMAT=materialized_path2 bundle exec rake test
• FORMAT=materialized_path3 bundle exec rake test
• FORMAT=array DB=pg bundle exec rake test
• ANCESTRY_COLUMN=family_tree bundle exec rake test