Replace id prefix text searches with binary index as for dataclips

User story

There are some searches in the app that allows filtering by id prefixes (history page and now on the dataclips selection with the new manual run).

We can save some db cycles filtering by binary id intervals instead of using like operator converting the id to text.

Details

Postgres performs a Parallel Sequential Scan when using the like operator while it can fully make use of the index when using binary intervals. This happens not only for uuid prefixes with odd number of chars but for any id prefix.

Implementation notes

Here is the dynamic filter: https://github.com/OpenFn/lightning/blob/9e255a660f7f3c6df36677bc073c955f9d47e36d/lib/lightning/invocation.ex#L72-L79

The case of prefixes of odd number of chars (1, 3, 5...) requires an additional filtering once 2 hex chars compose 1 byte https://github.com/OpenFn/lightning/pull/3196/commits/9e255a660f7f3c6df36677bc073c955f9d47e36d.

Release notes

User acceptance criteria

Start with the dynamic filter in lib/lightning/invocation.ex at lines 72-79, then compare the existing dataclips implementation and the approach referenced in pull request 3196. Trace the history-page and dataclips prefix searches, including odd-length prefixes, and verify that matching behavior and indexed query plans are preserved.