feat(compute): add dictionary_encode

[fallintoplace]

Rationale for this change

Arrow-Go currently has a vector hash kernel for unique, but not for dictionary encoding. The existing memo table already assigns stable positions to distinct values, so dictionary encoding can build on that state and provide a useful output form for repeated values.

What changes are included in this PR?

• Add the dictionary_encode compute function and DictionaryEncodeOptions.
• Support primitive, binary, temporal, decimal, interval, fixed-size binary, and boolean input types covered by the existing vector hash kernels.
• Use int32 dictionary indexes and preserve first-seen dictionary order.
• Mask nulls by default. With NullEncodingEncode, null is included in the dictionary and encoded as a regular index.
• Keep one shared dictionary across chunked input while returning one index per input value.
• Pass through already dictionary-encoded input instead of creating a nested dictionary type.
• Register boolean inputs in the shared vector hash kernels, which also adds unique(bool) support while preserving its existing null-as-a-distinct-value behavior.
• Correct the unique documentation to state that null is treated as a distinct value.

Are these changes tested?

• go test ./arrow/compute/... -count=1
• Added coverage for repeated and all-null values in both null modes, chunked input, dictionary input, dictionary order, output cardinality, boolean unique and dictionary encoding, sliced boolean input, and memo-table resizing.

Are there any user-facing changes?

Yes. This adds the public compute.DictionaryEncode function, DictionaryEncodeOptions, and the NullEncodingMask and NullEncodingEncode options. It also adds boolean input support to unique through the shared vector hash registration.