add support to log rtt

[KershawChang]

Maybe it's worth to also log RTT.

[martinthomson]

Logging on packet send rather than when we update the value (which is when we receive an ACK) seems like we'll log more often than is necessary.

[codecov-commenter]

Codecov Report

Attention: 12 lines in your changes are missing coverage. Please review.

Comparison is base (64fb41f) 87.63% compared to head (1cd1543) 87.54%.
Report is 15 commits behind head on main.

Files	Patch %	Lines
neqo-transport/src/rtt.rs	0.00%	10 Missing ⚠️
neqo-transport/src/cc/classic_cc.rs	77.77%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1522      +/-   ##
==========================================
- Coverage   87.63%   87.54%   -0.09%     
==========================================
  Files         117      118       +1     
  Lines       38323    38497     +174     
==========================================
+ Hits        33584    33704     +120     
- Misses       4739     4793      +54     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[KershawChang]

Logging on packet send rather than when we update the value (which is when we receive an ACK) seems like we'll log more often than is necessary.

Agreed. We should only log when it's changed.

[KershawChang]

Logging on packet send rather than when we update the value (which is when we receive an ACK) seems like we'll log more often than is necessary.

I've decided to log rtts in on_packets_acked. The main reason is that there could be multiple ClassicCongestionControl objects in a log and we need to link rtt information and the other information in ClassicCongestionControl together.
Please take another look. Thanks.

[martinthomson]

A simplification might be in order, but this is OK otherwise.

[martinthomson]

Any reason why this can't be passed &RttEstimate? The function is called from Path::on_packets_acked(), so I think that is much easier. Then your logging calls might just invoke the Debug implementation on RttEstimate (which might need to be improved, as it uses the default).

[martinthomson]

Your tests can probably build a const RTT_ESTIMATE.

[KershawChang]

Thanks. This is neat.
I'll do this.

[KershawChang]

I've addressed the comment again.
Please take another look. Thanks!

[martinthomson]

@larseggert, do you want to review this? If not, would you mind merging it?

[larseggert]

LGTM, but suggest renaming rtts -> rtt_est for clarity.

[KershawChang]

I assume I can just merge this PR, since the change after last review was addressing some nits.