Migrate to JDK 11 client or AHC5

[berezovskyi]

https://docs.oracle.com/en/java/javase/11/docs/api/java.net.http/java/net/http/HttpClient.html

Add support for HTTP2 and eliminates an unnecessary dependency.

[berezovskyi]

I am pretty sure we won't get it into Lyo 5.x (also for the purpose of not introducing more breaking incompatibilities) but here is some info I have been collecting:

• JDK 11 client does not have support for compression out of the box, which is tablestakes in modern HTTP comms: https://stackoverflow.com/questions/53502626/does-java-http-client-handle-compression
• There is a really nice wrapper https://mizosoft.github.io/methanol/ but this brings a question on how much are we gaining by removing Apache HttpClient just to introduce another dependency. Now, this may increase performance but need to think this through.
• JDK 11 client introduces new interfaces for (de-)serialization: https://docs.oracle.com/en/java/javase/11/docs/api/java.net.http/java/net/http/HttpRequest.BodyPublishers.html and the aforementioned client wrapper offers neat extension https://mizosoft.github.io/methanol/api/latest/methanol/com/github/mizosoft/methanol/MoreBodyHandlers.html

Alternatively, we should consider migrating from AHC 4.5 to AHC 5.0. AHC5 supports HTTP2 and there is a 4.5 to 5 migration guide with performance in mind. Feature comparison: https://www.mocklab.io/blog/which-java-http-client-should-i-use-in-2020/. AHC5 seems to win in every way except for the WebSocket support. I think that this is where we can use JDK 11 client as the communication modes are sufficiently different for HTTP req/resp and WebSocket: https://usoar.es/posts/consuming-websocket-with-java-http-client/.

More important could be to shift towards a more complex async model. We can safely make at least 6 requests to a server in parallel, IE8 in 2009 had such a limit. This is already logged as LYO-169.

[berezovskyi]

Interesting discussion on https://forum.open-services.net/t/rmsample-connectionclosedexception-exception/651/5:

1. Properly consuming responses is not a solved problem yet, migrating to Jakarta Restful-WS 2.1 would allow us to use try-with-resources in Java, which is less error-prone.
2. AsyncHttpClient from the comparison above is not going to support HTTP/2 ever as its semi-abandoned.
3. Java new default client also doesn't support HTTP forms requests, which we sadly need due to Jazz forms auth. Maybe not a big deal but adds to the lack of OOB compression support.

I'd say it leaves us with AHC 5 or OkHttp. OkHttp has Android support while I suspect that our AHC 4 code will be easier to migrate to AHC 5. I therefore think we should zero in on AHC 5 (with async in mind due to how spectacular #169 is with 50+ rps on a slow laptop against an OSLC adaptor with no perf tuning whatsoever).