Optimize ts v2 pool

[SungJin1212]

This PR improves V2 write request memory pooling by preventing string pointer leaks, adding explicit string clearing in ReuseWriteRequestV2 to prevent memory leaks when reusing pooled requests, and removing redundant exemplar label cleanup from ReuseTimeseriesV2.

Which issue(s) this PR fixes:
Fixes #

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[CharlieTLe]

The V2 pool was introduced in 918aa41 and shipped in v1.20.0, so the missing symbol clearing is a bug in a released version. Should a [BUGFIX] CHANGELOG entry be added?

[CharlieTLe]

Three symbols are appended ("", "name", "test") but only indices 0 and 1 are checked. Should index 2 be asserted as well?

Suggested change

	assert.Equal(t, "", underlyingArray[1])
	}
	assert.Equal(t, "", underlyingArray[1])
	assert.Equal(t, "", underlyingArray[2])
	}

[CharlieTLe]

Could this test flake since sync.Pool doesn't guarantee Get returns the same object that was just Put? If the GC runs between the ReuseWriteRequestV2 and PreallocWriteRequestV2FromPool calls, a fresh object would be returned and the backing array assertions would be skipped.

Would it be more robust to capture a reference to the backing array before the reuse call, so the clearing can be verified directly without depending on the pool?

// Capture backing array before reuse
backingArray := req.Symbols[:cap(req.Symbols)]

ReuseWriteRequestV2(req)

// Verify clearing directly on the backing array
for i, s := range backingArray[:3] {
    assert.Equalf(t, "", s, "symbol at index %d not cleared", i)
}

[SungJin1212]

I fixed it at the latest commit to check if the symbol table is reset without GC.
Thanks for the comment!