WIP: Only read relevant parts of proguard files into memory

[loewenheim]

Depends upon #1487 and https://github.com/getsentry/rust-proguard/pull/41.

This does a few things:

• When we download a proguard file, we create an index from it (see https://github.com/getsentry/rust-proguard/pull/41), zip the file and the index together, and persist them that way.
• When we load a proguard file into memory, we always do so with respect to a class name. We look up the relevant byte range in the index and only create a mapper from that part. Since we want to hold different data on disk and in memory, we need the change from #1487.

This change includes a cache version bump because we now store zip archives instead of just proguard files.

[loewenheim]

Superseded by #1491 .