Allow HTTP/2 max concurrent stream configuration

[tomMoulard]

What does this PR do?

This PR allows configuring the max concurrent stream value of HTTP/2 entrypoint.

Motivation

Fixes #8768

More

• Added/updated tests
• Added/updated documentation

[rtribotte]

LGTM 👌

[tomMoulard]

Adding the bot/no-merge label as there are issues with secured connections

[tomMoulard]

The issue should be fixed, removing the bot/no-merge label

[mpl]

Summary, explaining why we're delaying on that PR.
we're not seeing any problem at the moment, i.e. the use of x/net seems legitimate for this case, and how we used it seems correct.
However:

• it seems we'll have to keep x/net even more conscientiously updated - at least at every minor Go update, and maybe we have to think more whether that is a cost we are willing to incur
• we're still worried we might be forgetting something, and given we're a reverse-proxy we can't really afford to make mistakes in this area
• and independently, we're realizing that maybe for the h2c case, the http2.Server should have been (even before this PR) initialized with a priority write scheduler, instead of letting it default with a random write scheduler.

TL;DR: we'd like to think about it a bit more, and not rush it for 2.7.

[kevinpollet]

LGTM

[MaxThom]

Hey @tomMoulard, big thanks to you and all your colleagues.
Its working beautifully! No more node ports everywhere :P