Merge query cache and query state

[zetanumbers]

So basically we have query cache and query state separate "hash" tables. Cache tracks completed queries while state tracks unfinished queries. Thus we are required to a double check lookup while updating one of them like in try_execute_query and we do multiple hash table lookups while executing a query since cache/state entry might have moved due to rehashing. This PR merges query cache with query state and stores query cache entries in an arena for its stable memory location.

However my manually written slop code currently breaks query cycle detection, but I know how to fix it.

More details in #t-compiler/query-system > Merging query cache and query state

[zetanumbers]

And there're the benchmark results for 4 threads on my MacBook Pro M1 Pro:

Benchmark	Before	After
	Time	Time	%
🟣 regex:check	0.2070s	0.2031s	💚 -1.90%
🟣 regex:check:initial	0.2437s	0.2430s	-0.28%
🟣 regex:check:unchanged	0.1195s	0.1192s	-0.25%
🟣 syn:check	0.3599s	0.3546s	💚 -1.48%
🟣 syn:check:initial	0.4273s	0.4204s	💚 -1.63%
🟣 syn:check:unchanged	0.2503s	0.2473s	💚 -1.21%
🟣 clap:check	0.3343s	0.3274s	💚 -2.07%
🟣 clap:check:initial	0.3885s	0.3804s	💚 -2.10%
🟣 clap:check:unchanged	0.1589s	0.1582s	-0.46%
🟣 hyper:check	0.0757s	0.0731s	💚 -3.42%
🟣 hyper:check:initial	0.0913s	0.0882s	💚 -3.44%
🟣 hyper:check:unchanged	0.0545s	0.0548s	0.53%
Total	2.7109s	2.6694s	💚 -1.53%
Summary	1.0000s	0.9852s	💚 -1.48%

[rust-log-analyzer]

The job x86_64-gnu-miri failed! Check out the build log: (web) (plain enhanced) (plain)

Click to see the possible cause of the failure (guessed by this bot)