Improve performance of the History page queries

Coming off #968.

We have several places where performance of the History page query is hampered, here is a list of things I noticed:

1. We should not cast columns when searching, for example both log_lines and the dataclip tables have their columns cast into varchars - this is very expensive.
2. Finding the last run by joining on finished_at is exponentially less efficient depending on how many attempts & runs there are.

Proposed solutions

• #1898
• #1899
• Add index to attempt_runs for attempt_id and/or run_id, currently the compound index doesn't appear to be used for these queries.

These solutions are distinct (and should be done regardless) from taking another approach which is creating aggregate tables

1. Create aggregate (or columns) that get updated when an attempt/run is created or updated. Removing the need to join and filter Runs by exit code and finished_at.

References:

invocation.ex:417 - list_work_orders_for_project_query/2

Start at invocation.ex:417 and trace list_work_orders_for_project_query/2, then review the remaining unchecked work in #1899 and the proposed attempt_runs index. Done means the History page queries avoid the listed column casts and inefficient finished_at joins, with the query behavior and performance improvements verified against the stated cases.