Optimize column reader by bitmap counting definition levels

[aryan-cloud]

Rationale for this change

Fuse definition level counting into a bitmap-based pass to reduce extra memory traversal and improve cache/SIMD efficiency, addressing the TODO in column_reader.cc.

What changes are included in this PR?

Replace std::count with a bitmap/PopCount helper for counting max definition levels in ReadLevels.

Are these changes tested?

yes
cpp/build-asan/debug/parquet-encoding-test --gtest_filter=*
cpp/build-asan/debug/parquet-arrow-internals-test --gtest_filter=*

Are there any user-facing changes?

No.

[github-actions]

Thanks for opening a pull request!

If this is not a minor PR. Could you open an issue for this pull request on GitHub? https://github.com/apache/arrow/issues/new/choose

Opening GitHub issues ahead of time contributes to the Openness of the Apache Arrow project.

Then could you also rename the pull request title in the following format?

GH-${GITHUB_ISSUE_ID}: [${COMPONENT}] ${SUMMARY}

or

MINOR: [${COMPONENT}] ${SUMMARY}

See also:

• Other pull requests
• Contribution Guidelines - Contributing Overview

[HuaHuaY]

Suggested change

	int64_t offset = 0;
	for (; offset + 64 <= num_levels; offset += 64) {
	const uint64_t bitmap =
	internal::GreaterThanBitmap(levels + offset, 64, rhs);
	count += static_cast<int64_t>(bit_util::PopCount(bitmap));
	}
	if (offset < num_levels) {
	const uint64_t bitmap =
	internal::GreaterThanBitmap(levels + offset, num_levels - offset, rhs);
	count += static_cast<int64_t>(bit_util::PopCount(bitmap));
	}
	for (int64_t offset = 0; offset < num_levels; offset += 64) {
	const int64_t chunk_size = std::min<int64_t>(64, num_levels - offset);
	const uint64_t bitmap =
	internal::GreaterThanBitmap(levels + offset, static_cast<int>(chunk_size), rhs);
	count += static_cast<int64_t>(bit_util::PopCount(bitmap));
	}

I suggest simplifying the two loops into one.