introduce locking to manage access & download of archive index

[syphar]

This comes for

1. some race conditions with index files I want to solve. This will get more with other changes I have planned.
2. the ability to later add cleanup for the local caches, without affecting requests, if we need that

Also, every now and then I manually wipe the cache, and let the server re-download the files when they are accessed.

When I do that, we see regular sqlite "database image is malformed" errors, likely because of a race condition between the index download, rename and continous requests happening. All the time following requests on the same URLs that had these errors work fine.

The shared RwLock protects us here. Downside is that requests will queue up until the first request downloaded the archive index, and then continue. But IMO this is good behaviour.

[GuillaumeGomez]

Although I think the read/write locks API definitely should be improved, we didn't yet come up with a better API. I'll try to dive into it later on.