core: Switch to FxHash to improve performance

[penberg]

The default Rust hash map is slow for integer keys. Switch to FxHash instead to reduce executed instructions for, for example, throughput benchmark.

Before:

penberg@turing:~/src/tursodatabase/turso/perf/throughput/turso$ perf stat ../../../target/release/write-throughput --threads 1 --batch-size 100 --compute 0 -i 10000
Turso,1,100,0,106875.21

 Performance counter stats for '../../../target/release/write-throughput --threads 1 --batch-size 100 --compute 0 -i 10000':

          2,908.02 msec task-clock                       #    0.310 CPUs utilized
            30,508      context-switches                 #   10.491 K/sec
               261      cpu-migrations                   #   89.752 /sec
               813      page-faults                      #  279.572 /sec
    20,655,313,128      instructions                     #    1.73  insn per cycle
                                                  #    0.14  stalled cycles per insn
    11,930,088,949      cycles                           #    4.102 GHz
     2,845,040,381      stalled-cycles-frontend          #   23.85% frontend cycles idle
     3,814,652,892      branches                         #    1.312 G/sec
        54,760,600      branch-misses                    #    1.44% of all branches

       9.372979876 seconds time elapsed

       2.276835000 seconds user
       0.530135000 seconds sys

After:

penberg@turing:~/src/tursodatabase/turso/perf/throughput/turso$ perf stat ../../../target/release/write-throughput --threads 1 --batch-size 100 --compute 0 -i 10000
Turso,1,100,0,108663.84

 Performance counter stats for '../../../target/release/write-throughput --threads 1 --batch-size 100 --compute 0 -i 10000':

          2,838.65 msec task-clock                       #    0.308 CPUs utilized
            30,629      context-switches                 #   10.790 K/sec
               351      cpu-migrations                   #  123.650 /sec
               818      page-faults                      #  288.165 /sec
    19,887,102,451      instructions                     #    1.72  insn per cycle
                                                  #    0.14  stalled cycles per insn
    11,593,166,024      cycles                           #    4.084 GHz
     2,830,298,617      stalled-cycles-frontend          #   24.41% frontend cycles idle
     3,764,334,333      branches                         #    1.326 G/sec
        53,157,766      branch-misses                    #    1.41% of all branches

       9.218225731 seconds time elapsed

       2.231889000 seconds user
       0.508785000 seconds sys

[nyrkio]

Nyrkiö Report for Commit: dfab8c4

No performance changes detected.
Remember that Nyrkiö results become more precise when more commits are merged.
So please check again in a few days.

0 changes / 130 tests & 302 metrics.