perf: classify encoder table entries once

[pelletier]

Splitting #1088: this PR carries one optimization technique so its impact and review surface stay isolated. First of the encoder chain.

What

Deciding whether an entry is a key-value, a table, or an array of tables requires resolving the value's type — and for arrays, resolving the type of every element. That decision was re-derived by each encoding pass (and a third time by SetOmitEmptySuperTables's super-table check), up to three times per entry.

Entries now carry an entryClass, computed at most once and cached on the entry: the first pass classifies while emitting key-values, the second pass routes on the stored class, and onlySubTables reuses whatever the entries already know (classOf computes lazily for the entries it sees first). entryIsTable disappears.

Marshaler-shaped entries keep their rawShape routing and are assigned their class by it, so the opt-in unstable.Marshaler semantics are untouched.

Impact

Benchmarked on a dedicated linux/amd64 spot VM (t2d), go1.26.4, interleaved A/B vs the base of this PR, benchstat over 10 samples per side:

• sec/op geomean: -2.45%
• Marshal/HugoFrontMatter: -9.85%
• Marshal/ReferenceFile/struct: -9.63%
• RealWorldViperWrite: -8.13%
• Marshal/SimpleDocument/map: -7.63%
• Marshal/SimpleDocument/struct: -7.23%
• Marshal/ReferenceFile/map: -7.02%

Full benchstat (sec/op, allocs/op)

sec/op

UnmarshalDataset/config  10.24m ± 1%  10.27m ± 1%  ~ (p=0.529 n=10)
UnmarshalDataset/canada  23.26m ± 1%  23.16m ± 1%  ~ (p=0.912 n=10)
UnmarshalDataset/citm_catalog  14.91m ± 1%  14.79m ± 1%  ~ (p=0.143 n=10)
UnmarshalDataset/twitter  3.572m ± 0%  3.566m ± 0%  ~ (p=0.075 n=10)
UnmarshalDataset/code  33.32m ± 1%  33.45m ± 1%  ~ (p=0.631 n=10)
UnmarshalDataset/example  77.10µ ± 0%  77.01µ ± 2%  ~ (p=0.529 n=10)
Unmarshal/SimpleDocument/struct  332.0n ± 1%  328.6n ± 3%  ~ (p=0.159 n=10)
Unmarshal/SimpleDocument/map  423.1n ± 1%  423.4n ± 1%  ~ (p=0.542 n=10)
Unmarshal/ReferenceFile/struct  39.04µ ± 2%  38.87µ ± 2%  ~ (p=0.853 n=10)
Unmarshal/ReferenceFile/map  31.21µ ± 1%  31.15µ ± 1%  ~ (p=0.853 n=10)
Unmarshal/HugoFrontMatter  5.602µ ± 1%  5.615µ ± 1%  ~ (p=0.448 n=10)
Marshal/SimpleDocument/struct  397.4n ± 1%  368.7n ± 3%  -7.23% (p=0.000 n=10)
Marshal/SimpleDocument/map  592.4n ± 1%  547.2n ± 2%  -7.63% (p=0.000 n=10)
Marshal/ReferenceFile/struct  20.44µ ± 1%  18.47µ ± 4%  -9.63% (p=0.000 n=10)
Marshal/ReferenceFile/map  38.89µ ± 2%  36.16µ ± 3%  -7.02% (p=0.000 n=10)
Marshal/HugoFrontMatter  7.941µ ± 3%  7.158µ ± 2%  -9.85% (p=0.000 n=10)
RealWorldContainerdConfig  39.67µ ± 1%  39.26µ ± 2%  ~ (p=0.089 n=10)
RealWorldViperRead  8.241µ ± 1%  8.210µ ± 0%  ~ (p=0.160 n=10)
RealWorldViperWrite  12.67µ ± 3%  11.64µ ± 3%  -8.13% (p=0.000 n=10)
RealWorldHugoFrontMatterBatch  93.86µ ± 2%  92.58µ ± 1%  -1.36% (p=0.000 n=10)
RealWorldGitleaksRules  63.43µ ± 3%  63.74µ ± 1%  ~ (p=0.971 n=10)
RealWorldPyproject  14.85µ ± 1%  14.71µ ± 1%  -0.96% (p=0.029 n=10)
RealWorldGolangciStrict  11.73µ ± 1%  11.75µ ± 1%  ~ (p=0.753 n=10)
geomean  46.26µ  45.13µ  -2.45%

allocs/op

UnmarshalDataset/config  70.19k ± 0%  70.19k ± 0%  ~ (p=1.000 n=10) ¹
UnmarshalDataset/canada  223.2k ± 0%  223.2k ± 0%  ~ (p=1.000 n=10) ¹
UnmarshalDataset/citm_catalog  49.98k ± 0%  49.98k ± 0%  ~ (p=1.000 n=10) ¹
UnmarshalDataset/twitter  15.78k ± 0%  15.78k ± 0%  ~ (p=1.000 n=10) ¹
UnmarshalDataset/code  129.5k ± 0%  129.5k ± 0%  ~ (p=1.000 n=10) ¹
UnmarshalDataset/example  399.0 ± 0%  399.0 ± 0%  ~ (p=1.000 n=10) ¹
Unmarshal/SimpleDocument/struct  2.000 ± 0%  2.000 ± 0%  ~ (p=1.000 n=10) ¹
Unmarshal/SimpleDocument/map  5.000 ± 0%  5.000 ± 0%  ~ (p=1.000 n=10) ¹
Unmarshal/ReferenceFile/struct  83.00 ± 0%  83.00 ± 0%  ~ (p=1.000 n=10) ¹
Unmarshal/ReferenceFile/map  194.0 ± 0%  194.0 ± 0%  ~ (p=1.000 n=10) ¹
Unmarshal/HugoFrontMatter  54.00 ± 0%  54.00 ± 0%  ~ (p=1.000 n=10) ¹
Marshal/SimpleDocument/struct  4.000 ± 0%  4.000 ± 0%  ~ (p=1.000 n=10) ¹
Marshal/SimpleDocument/map  4.000 ± 0%  4.000 ± 0%  ~ (p=1.000 n=10) ¹
Marshal/ReferenceFile/struct  4.000 ± 0%  4.000 ± 0%  ~ (p=1.000 n=10) ¹
Marshal/ReferenceFile/map  91.00 ± 0%  91.00 ± 0%  ~ (p=1.000 n=10) ¹
Marshal/HugoFrontMatter  20.00 ± 0%  20.00 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldContainerdConfig  144.0 ± 0%  144.0 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldViperRead  65.00 ± 0%  65.00 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldViperWrite  33.00 ± 0%  33.00 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldHugoFrontMatterBatch  820.0 ± 0%  820.0 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldGitleaksRules  259.0 ± 0%  259.0 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldPyproject  142.0 ± 0%  142.0 ± 0%  ~ (p=1.000 n=10) ¹
RealWorldGolangciStrict  48.00 ± 0%  48.00 ± 0%  ~ (p=1.000 n=10) ¹
geomean  211.1  211.1  +0.00%
¹ all samples are equal

🤖 Generated with Claude Code