Refactor count summary for flexibility + more tests

[olorin]

On top of #99. diff

• Add a "summarize" step for the row counts so we can perform additional transformations on the final accumulated state before serializing it (no schema changes in this PR).
• Some associative/commutative properties for appropriate things.
• Benchmark for numeric stuff.

[olorin]

Benchmark bench: RUNNING...
benchmarking decoding/decode/conduit+attoparsec-bytestring/1000
time                 285.6 ms   (282.3 ms .. 289.9 ms)
                     1.000 R²   (0.999 R² .. 1.000 R²)
mean                 291.5 ms   (289.9 ms .. 293.6 ms)
std dev              2.002 ms   (503.9 μs .. 2.607 ms)
variance introduced by outliers: 16% (moderately inflated)

benchmarking field-parsing/parseField/200
time                 83.93 μs   (82.00 μs .. 85.31 μs)
                     0.993 R²   (0.988 R² .. 0.996 R²)
mean                 76.76 μs   (74.37 μs .. 79.26 μs)
std dev              8.397 μs   (7.690 μs .. 9.689 μs)
variance introduced by outliers: 85% (severely inflated)

benchmarking folds/updateSVParseState/1000
time                 154.7 ms   (146.7 ms .. 162.8 ms)
                     0.997 R²   (0.994 R² .. 1.000 R²)
mean                 143.2 ms   (136.5 ms .. 147.9 ms)
std dev              7.793 ms   (5.631 ms .. 10.56 ms)
variance introduced by outliers: 13% (moderately inflated)

benchmarking folds/hashText/1000
time                 112.1 μs   (97.38 μs .. 124.4 μs)
                     0.955 R²   (0.946 R² .. 0.994 R²)
mean                 107.8 μs   (101.5 μs .. 122.3 μs)
std dev              29.12 μs   (14.35 μs .. 52.78 μs)
variance introduced by outliers: 97% (severely inflated)

benchmarking folds/updateTextCounts/1000
time                 26.25 ms   (21.11 ms .. 31.49 ms)
                     0.911 R²   (0.810 R² .. 0.973 R²)
mean                 32.98 ms   (30.82 ms .. 35.79 ms)
std dev              5.653 ms   (3.914 ms .. 7.671 ms)
variance introduced by outliers: 69% (severely inflated)

benchmarking numerics/updateNumericState/10000
time                 6.135 ms   (5.956 ms .. 6.279 ms)
                     0.992 R²   (0.987 R² .. 0.996 R²)
mean                 6.669 ms   (6.521 ms .. 6.961 ms)
std dev              592.5 μs   (469.3 μs .. 738.9 μs)
variance introduced by outliers: 52% (severely inflated)

benchmarking numerics/combineMeanDevAcc/10000
time                 391.8 μs   (375.0 μs .. 410.0 μs)
                     0.987 R²   (0.978 R² .. 0.996 R²)
mean                 365.2 μs   (358.1 μs .. 376.5 μs)
std dev              29.51 μs   (17.76 μs .. 43.37 μs)
variance introduced by outliers: 69% (severely inflated)

benchmarking numerics/combineNumericState/10000
time                 483.8 μs   (464.9 μs .. 508.1 μs)
                     0.988 R²   (0.982 R² .. 0.994 R²)
mean                 500.3 μs   (486.8 μs .. 512.8 μs)
std dev              44.17 μs   (35.98 μs .. 51.68 μs)
variance introduced by outliers: 72% (severely inflated)

Benchmark bench: FINISH

[charleso]

I'm assuming SVParseState can't just be renamed/re-used as RowCountSummary as they will diverge at some point?

[olorin]

Yep, see PR #102 - the numeric state will be added, and needs to be finalized before it's serialized.

[charleso]

👍 for everything not in #99