fix ~10x performance regression by reverting a change

[LGriffioen]

change was introduced with zip64 support

Fixes #311

Changes proposed in this PR

• reverted b3ca1e6#diff-d829f12a7b0fa38c616a3397c9ea1464287b0d0097dee344876fff88b4b28e39L316

Tests performed

• performance of loading 10,000 files that are zipped is ~10x faster than development, and roughly where it was before the Zip64 changes
• I'm not familiar enough with Zip64 to be able to test what impacts this change has on that functionality

Further info for the reviewer

• Performance with large numbers of files is critical for us. This 10x improvement is taking a 20s operation down to under 2s

Open Issues

[LGriffioen]

@weichsel @Ckitakishi i just rebased this, any chance either of you could take a look? thanks!

[weichsel]

Thanks for keeping this up to date.
I checked the changes, and they look good. They follow the ZIP specification, and since they only affect parsing of the standard EOCD record, there shouldn’t be any side effects or regressions related to ZIP64 support. This should also address the overall EOCD parsing performance regression. 🙏

[LGriffioen]

thanks for taking a look!