Store runtime benchmark results into SQLite

[Kobzol]

Currently, each iteration of a runtime benchmark is stored with a fresh collection ID. This is something that we may want to change, although there is currently a unique index across (pstat_series, artifact_id and collection_id), so we cannot easily associate multiple metric sets with a single collection ID.

[Kobzol]

ping @Mark-Simulacrum

[Mark-Simulacrum]

Didn't review in too much detail but this looks fine to me at a high level. I imagine we can iterate on the specifics (e.g., whether we need a separate table for these benchmarks) over time - until we add Postgres support and start collecting in production we can throw the tables away easily without worrying about migrations etc.