API doing a ReadAll() on input is not reliable due to memory constraints

[phish3y]

No description provided.

[frantjc]

we pretty much need to just take their content-type header and assume they're correct about what they're sending, right? no more validation.

[frantjc]

imo this would speed up upload times a lot, I have a feeling uploads are slow due to resource constraint differences between our cluster and our monsterous workstations with nothing else running on them

[phish3y]

@frantjc I'm willing to drop the content type verification in the API layer to support better performance and reliability. Let it fail in the task if the input is BS

[frantjc]

There's also some code in objectstore/s3_volume.go that is not properly implementing io.Reader which would save on memory when streaming big files from s3 I think.