docs: clarify operational differences between throttling and postponing cut-over (#1775)

[soepic1]

Fixes #1775

Description

This PR improves the operational documentation for large-scale migrations (100M+ rows / multi-hour executions):

1. Throttling vs. Postponing Cut-Over: Adds a clear comparison table and best-practice operational workflow detailing why --postpone-cut-over-flag-file is preferred over long-term --throttle-additional-flag-file during scheduled maintenance windows (preventing MySQL wait_timeout disconnects while maintaining real-time binlog sync).
2. UUID / String Primary Key Guidance: Documents progress percentage behavior (e.g., overshoot past 100% and ETA: due) on tables with lexicographical/UUID primary keys and explains how operators can verify boundary progress.

Related Issue

Closes #1775

[soepic1]

Hi @timvaillancourt,
I've just updated this branch to resolve the conflicts with master.
It looks like the CI workflows are awaiting approval to run. Would it be possible to approve them so the automated checks can complete?
Happy to make any changes based on the results. Thank you!

[soepic1]

Hi @timvaillancourt,

I've investigated the CI test failure. It looks like the copy-retries-exhausted test is failing on MariaDB 10.11.

The error log indicates that the test failed because the gh-ost execution succeeded when it was expected to fail:

ERROR copy-retries-exhausted execution was expected to exit on error but did not.
Since my PR only modifies documentation, this seems like it might be a flaky test.

Is there a way to re-run the failed check? I'm happy to help in any way I can. Thanks!