fix(queryset): hydrate select_related by field order (Fixes #1902)

[vyrnsynx]

Description

Postgres truncates identifiers to 63 bytes. Tortoise select_related JOINs build aliases such as table__long_relation__nested.field and execute_select used to recover field names by splitting those aliases on .. After truncation this raises IndexError (the . is gone) or hydrates the wrong attribute (AttributeError on long OneToOne / column names).

Related columns were already sliced by position via select_related_idx. This change records the ordered field names for each related group when the query is built and zips them with row values, so field identity no longer depends on untruncated aliases. SQLite and MySQL keep the same SELECT shape; only hydration mapping changes.

Motivation and Context

Fixes tortoise#1902

How Has This Been Tested?

• New regression tests for nested long FK aliases, long OneToOne model + column names, and .only() + select_related
• Existing select_related / .only() / relations tests on SQLite
• Helper test that positional row access survives duplicate truncated keys

Checklist:

• My code follows the code style of this project.
• I have added the changelog accordingly.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.
• All new and existing tests passed.

  

[vyrnsynx]

Superseded by tortoise#2270