Attempted solution to db pool deadlock issue

[hardbyte]

tl;dr: Dependencies and path operations defined as functions (def only) are run in an anyio threadpool. db.execute is a blocking call, causing all the workers in the pool to block in the path operation function. This prevents dependency generators from invoking their finally block, thereby preventing SQLAlchemy connections from being released.
fastapi/fastapi#3205 (comment)

The proper solution is likely to use asyncio compatible database - which is coming in Sqlalchemy... but is also a huge rewrite of the code base.

A lovely hack of a solution was suggested in fastapi/full-stack-fastapi-template#104 (comment) which is what this PR adds. Don't merge until we see how it goes...

[hardbyte]

Comparing against the baseline in #139 this hack seems to help. I haven't repeated the experiments but this had no errors up to 85 users @ 35 requests per second. It failed when the 3rd container came up, and exhausted the available db connections.

Suggest we merge this and establish as a new baseline.