Avoid double-fetching of chunk mappings on retries in file downloads

[clostao]

Problem:

Due to the fragility of read operations over the subspace-gateway, the package @autonomys/auto-files that is an interface for handling chunked file downloads implements a retries approach for handling the likely errors during a big file download.

When a retry is submitted for a certain chunk, all the objects whose start is within that object are fetched again, making the subspace gateway to be fetching twice those object mappings.

A possible solution:

Implement a component in auto-files-gateway that would be responsible for avoiding this double fetching by deduplicating or caching the mapping requests for the same chunk/object ranges during retries.

[clostao]

Update: Retries are not an issue right now since timeouts solve this issue right now. This issue is still an issue when two users trigger the same file download.