feat: denormalize tree details with rollup table to improve summary performance

[gustavobtflores]

Problem

The tree details summary endpoint (e.g. tree/<commit_hash>/summary) is slow and heavy because it:

• Uses a single large query (get_tree_details_data) that joins builds, tests, incidents, and issues, returning one row per test.
• Fetches and transfers a large result set for commits with many tests.
• Performs all aggregation (status counts, arch/config summaries, issues, filters) in application code by iterating over those rows and deduplicating.
• This does not scale for trees with many tests and makes the summary endpoint slow and resource-intensive.