gh-156539: Validate duplicate ZIP members individually in testzip

[lpyu001]

Fix ZipFile.testzip() to pass each ZipInfo object directly to open() instead of reopening the member by filename.

When an archive contains duplicate filenames, filename-based lookup resolves to the last matching entry. As a result, testzip() can iterate over an earlier ZipInfo but actually reopen and validate a later duplicate, allowing corruption in the earlier member to go undetected.

Passing the ZipInfo directly preserves the identity of the member being iterated and ensures that every entry in filelist is validated individually.

This also avoids an unnecessary filename-to-ZipInfo lookup through NameToInfo and makes the implementation better match the existing ZipFile.open() API, which already accepts ZipInfo objects specifically to disambiguate duplicate members.

• Issue: ZipFile.testzip() Skips Earlier Duplicate Members #156539

[lpyu001]

Could you take a look at this PR when you have a chance? @pitrou

Since testzip() already iterates over the exact ZipInfo objects, I don't think it needs to look the member up again by name. Passing the ZipInfo directly preserves member identity, avoids an unnecessary lookup, and handles duplicate filenames correctly.

the relevant commit：
c534270