bodyReader: fall back to a different mirror when mid-stream reconnect fails

[QiWang19]

Problem

When bodyReader detects a connection drop mid-download (io.ErrUnexpectedEOF,
ECONNRESET), it reconnects to the same source using an HTTP Range request
to resume from the current offset. If that reconnect also fails, the download
fails

Context

The blob-level mirror fallback added in #845 handles failures at the
initial GetBlob request level, before any data flows. bodyReader handles
failures after data has started flowing.

This was pointed out during review of #845: #845 (review)

[There is a question whether bodyReader should also trigger a fallback to a different mirror… I think that would better be a different PR, and it might not even be necessary — OTOH there it might be worth just thinking a bit about what the code structure to enable that would look like.]

[mtrmac]

@QiWang19 From this repo’s point of view, I think deferring this part is basically fine; but from the perspective of users — vast majority of time spent fetching images is inside bodyReader. So I think not having a mirror fallback on such failures might significantly limit the benefit of the whole feature.