Efficient queries for connection list

[pmossman]

What

The web_backend connections list handler finds all connections within a workspace, and loops over each one, calling other handlers to build source reads, destination reads, and fetching the latest job info. Doing this once per connection is very slow and can require hundreds of queries for workspaces with many connections.

Instead, this PR performs all necessary queries up front and stores the information in maps that can then be referenced when building the response model for each connection. This should drastically reduce the number of database queries.

How

• Write new repository methods that fetch information given lists of connections/sources/etc.
• Write helpers that convert lists to maps for easy lookups
• Pass these maps into the methods that construct response items to eliminate queries inside the connection loop

Performance Testing

I did some comparison in dev between master and this branch for a workspace with ~250 active connections

• The code in master generates ~3.3k individual queries, and takes ~15-20 seconds. Here's an example span
• The code in this branch generates 11 queries, and this number is constant no matter how many connections are in the workspace. Here's an example span. The workspace loads in ~1-3 seconds instead of ~15-20 seconds, which is obviously a massive improvement.

Tagging @malikdiarra and @davinchia for review as they both have context on our repository layer and experience with making similar optimizations. Definitely looking for thoughts and feedback on the new queries and in-memory grouping code.

[davinchia]

Good points:

• we removed another few queries by combining the joins.
• this is so SO much more readable!

I'm curious what kind of performance improvement we have here.