GH-533: Adaptive Lossless Floating-Point (ALP) Encoding

[alamb]

This is a proposed implementation of the ALP described in

"ALP: Adaptive Lossless floating-Point Compression" (SIGMOD 2024, https://dl.acm.org/doi/10.1145/3626717

It is based on (largely a reformatted version of) @prtkgaur 's ALP Encoding Specification Google Doc

See rendered preview here: https://github.com/alamb/parquet-format/blob/alamb/alp/Encodings.md#adaptive-lossless-floating-point-adaptive_lossless_floating_point--10

Rationale for this change

This encoding has the following properties:

• Targets real-world floating-point (IEEE 754) data.
• It achieves higher compression ratios (close to ZSTD)
• Much faster to decompress than zstd (and other floating point algorithms)

See Mailing List Discussion: https://lists.apache.org/thread/tjtln1mmjqfoql1ls2dw9xpdk91r1909

Source ALP Results Document

(Todo summarize the mailing list discussion here)

What changes are included in this PR?

• Closes [Proposal] Add ALP encoding support in parquet file format #533

Do these changes have PoC implementations?

Yes

• C/C++: GH-48701: [C++][Parquet] Add ALPpd encoding arrow#48345