Bump zstd-seekable-format-go to v0.10.0

[SaveTheRbtz]

What changed

Bumps github.com/SaveTheRbtz/zstd-seekable-format-go/pkg from v0.8.0 to v0.10.0: https://github.com/SaveTheRbtz/zstd-seekable-format-go/releases/tag/pkg%2Fv0.10.0

This keeps sqlitezstd compiling against the stable v0.10 API while preserving the existing VFS behavior.

API migration

• Updated ZstdFile to store *seekable.Reader, matching the concrete Reader returned by seekable.NewReader in pkg: return concrete types SaveTheRbtz/zstd-seekable-format-go#264.
• Ran go mod tidy, which also picked up the v0.10 module's transitive dependency versions.

Validation

• go test -count=1 ./...

[SaveTheRbtz]

JFYI: there is also support for configurable cache on the zstd-seekable-format-go side, so you can remove up to 1K lines of code:

• Use seekable Sieve for frame caching SaveTheRbtz/sqlitezstd#1

There are multiple levels of cache removals in the PR, you can pick a subset of commits

• compressed blocks cache
• decompressed blocks cache
• http request cache

Having one cache benefits non-infinite memory scenarios where we do not want to duplicate the same data in 3 different caches.

I've ran basic perf tests with linux kernel sources sqlite database with fts5 index served over http. A single Sieve cache has the +/- same performance as 3 separate ones after we control for memory usage

[jtarchie]

I will take a look, thanks for the work.

I'm just working on loge to support some more efficient work reading over HTTP. So a product was driving out the feature. I'll see if I can use that to benchmark.

I see you have an upstream-upstream PR. Should I merge this first, then will you add the other one?