Nginx Config improvement

[zerox80]

Large sync runs still die after 1 000 HTTP/2 requests because the current nginx snippet in the docs only tweaks keepalive_timeout. As of nginx 1.24 the http2_max_requests/http2_idle_timeout directives are obsolete, so without explicit keepalive_requests the server keeps sending GOAWAY once 1 000 requests are processed. Desktop clients then abort with “PROPFIND response is not XML”.

Proposal
Extend the TLS vhost example in the docs (OpenCloud block) with:

keepalive_requests 100000;
keepalive_timeout 5m;
http2_max_concurrent_streams 512;
This keeps HTTP/2 connections alive for large TUS/PROPFIND batches and prevents the client errors. Reproduced on the current desktop beta with a large repo (Linux kernel); logs available on request.

[zerox80]

And maybe extend it with http2 :) So all desktop clients etc work well and fast

[micbar]

Fixed by #519