Introduce the (internal/db level) concept of "dynamic regions" for TIs

[ashb]

As mentioned in the first docs PR of this stack, and the precursor PR #74222,
we need a way of distinguishing between "interation 2" and "iteration 2 after
we already cleared and re-ran loop 2 but-these-are-totally-different-TIs, not
retries.", or more simply in the digram as 2' (said as "two prime").

Task loops need the same task_id to exist once per pass, and a clear needs to
put replacement task beside the work it has "retired" or superceded. A task
instance is identified by dag_id, task_id, run_id and map_index today, so
neither is possible: the second row would collide with the first as they both
have the same task_id and try_number.

A "dynamic region is one execution of a construct that creates task instances at
run time: a loop, or the expansion of a mapped task or task group. Each region
is a row recording which Dag run it belongs to, which node it executes (node_id,
the id of the task group or mapped task), which enclosing region and position
contains it and (so that we can have mapped tasks inside loops!), when a clear
replaced an earlier execution, which region it was forked from and where it
resumes (i.e. "clear iteration 2 onwards").

A region is a row rather than a position (i.e. it's own table, not just a colum
on TI) because a position can be reused/"overwritten" after a clear, and would
then stop saying which precise TI created the data. Regions are immutable for
the same reason; everything that changes stays on the task instances.

This change only adds the storage to hopefully make PR creation easier.
Nothing creates a region yet, and existing rows take the all-zero sentinel,
meaning outside any region, because rewriting existing rows on upgrade was
measured at production scale and rejected (4+ hrs! But I never let it run to
completion. Even 30mins is too long). XCom and rendered fields are keyed by
attempt UUID since the precursor ownership PR and so don't need a region column.
region_index is an ORM synonym of map_index for now.

Later changes will build upon this to make mapped expansions and loops create
regions, resolve and write task data by (region_id, region_index), add the task
loop gate that decides whether a loop runs another pass/iteration, and let a
clear replace only the selected work by forking a region instead of rewriting
history.

Downgrade refuses when rows differ only by region, because dropping the column
would otherwise merge distinct executions. The check scans task_instance, so
expect it to be slow on very large installations.

---

Was generative AI tooling used to co-author this PR?

• Yes (please specify the tool below)

---

• Read the Pull Request Guidelines for more information. Note: commit author/co-author name and email in commits become permanently public when merged.
• For fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
• When adding dependency, check compliance with the ASF 3rd Party License Policy.
• For significant user-facing changes create newsfragment: {pr_number}.significant.rst, in airflow-core/newsfragments. You can add this file in a follow-up commit after the PR is created so you know the PR number.