GH-47978: [C++][Parquet][CI] Add more compression codecs to fuzzing seed corpus

[pitrou]

Rationale for this change

1. Add more compression codecs to seed corpus
2. Tweak fuzz target to make fuzzing slightly faster (around ~30% locally according to my measurements), which will allow testing more mutations per day

Are these changes tested?

Not specifically by CI, but hopefully they will make fuzzing more efficient.

Are there any user-facing changes?

No.

• GitHub Issue: [C++][Parquet][CI] Add more compression codecs to fuzzing seed corpus #47978

[pitrou]

@github-actions crossbow submit fuzz

[github-actions]

Revision: efa842a

Submitted crossbow builds: ursacomputing/crossbow @ actions-9baaa59a9e

Task	Status
test-build-cpp-fuzz

[pitrou]

@github-actions crossbow submit fuzz

[github-actions]

Revision: 4bbdd6a

Submitted crossbow builds: ursacomputing/crossbow @ actions-001f65930b

Task	Status
test-build-cpp-fuzz

[pitrou]

@wgtmac @EnricoMi I would welcome a review on this.

[EnricoMi]

Nice! What about the other compressions (LZO, BZ2, LZ4_*)?

[pitrou]

These are all the compression codecs that Parquet C++ supports. BZ2 is not in the Parquet format, and LZO is closed source.

[EnricoMi]

These are all the compression codecs that Parquet C++ supports. BZ2 is not in the Parquet format, and LZO is closed source.

Fine with me, was just wondering since they are defined:

arrow/cpp/src/arrow/util/type_fwd.h

Lines 46 to 60 in f8a0902

	struct Compression {
	/// \brief Compression algorithm
	enum type {
	UNCOMPRESSED,
	SNAPPY,
	GZIP,
	BROTLI,
	ZSTD,
	LZ4,
	LZ4_FRAME,
	LZO,
	BZ2,
	LZ4_HADOOP
	};
	};

[EnricoMi]

LGTM!

[pitrou]

Fine with me, was just wondering since they are defined:

Well, some of them are supported by other parts of Arrow. For example you could read a BZ2-compressed CSV file.