Added some fault-tolerance to the build queue

[Kixiron]

Closes #628

Stops losing crates that are being added to the build queue when a crash occurs. Attempted to make add_crate_to_queue batched, but couldn't figure out how to with the current db driver

[Nemo157]

I just noticed the

add_crate_to_queue(&conn, &krate.name, &krate.version, priority).ok();
debug!("{}-{} added into build queue", krate.name, krate.version);

hidden in there, maybe this should be changed to

match add_crate_to_queue(&conn, &krate.name, &krate.version, priority) {
  Ok(_) => debug!("{}-{} added into build queue", krate.name, krate.version),
  Err(err) => error!("failed adding {}-{} into build queue: {}", krate.name, krate.version, err),
}

[Kixiron]

I realized that we could also add a prometheus hook for this, if it's something we think is important enough

[jyn514]

I realized that we could also add a prometheus hook for this, if it's something we think is important enough

I don't this is necessary, we already track the number of crates queued and we can look at the logs or /releases/queue if we want to know the exact crates.

[jyn514]

Looks great, thanks for working on this! Do you have any ideas on how to test this? I thought maybe we could delete the crate_priorities table in the test so that getting the connection and fetching the index succeed, but adding the crate fails.

[pietroalbini]

Why did you change the permissions of the docker entrypoint?

[Kixiron]

Fixed, something with Windows messed up

[jyn514]

Thanks for working on this!