PR #85 extended / v.1.2.4

[guruz]

FYI @zerox80

[guruz]

(This is still WiP, will put something on top still)

[zerox80]

has conflicts...

[zerox80]

I checked this against the OpenCloud backend implementation as well.

A couple of retry/resume edge cases look worth fixing before merge:

1. UploadFileFromContentUriWorker.ensureOriginalTusChecksum() reopens the original contentUri when a valid cached file already exists. For old/pre-checksum retries, the original SAF URI may be gone or permission-revoked while cachePath is still valid. In that case we should hash the cached file instead of failing with LocalFileNotFoundException.

2. In both upload workers, shouldTryTus is computed before clearing an old pending TUS state without SHA1. After clearTusState(), the stale shouldTryTus == true can still route the upload through a fresh TUS creation, even for small files or when capabilities no longer advertise TUS. Recomputing after clearing, or moving the clear before the decision, should avoid that.

Backend note: I initially wondered whether Upload-Checksum should be gated on tus_support.extension, but OpenCloud’s OCS capability currently advertises only creation,creation-with-upload while the actual TUS endpoint supports checksum. So sending the checksum as this PR does seems correct for OpenCloud.