Replace neqo_common::log with log crate

[mxinden]

neqo* currently uses neqo_common::log for all things logging. It is a wrapper around the log crate.

What is the benefit of using neqo_common::log over using the log crate directly? From what I can tell:

• neqo_common::log does not require explicit instantiation of the logger (e.g. via env_logger::init) but instead tries to initialize on each log invocation.
  

  neqo/neqo-common/src/log.rs

  Lines 84 to 90 in c209c43

  	#[macro_export]
  	macro_rules! log_invoke {
  	($lvl:expr, $ctx:expr, $($arg:tt)*) => ( {
  	::neqo_common::log::init(None);
  	::neqo_common::do_log!($lvl, "[{}] {}", $ctx, format!($($arg)*));
  	} )
  	}

• Uses elapsed time instead of absolute time
  

  neqo/neqo-common/src/log.rs

  Lines 69 to 71 in c209c43

  	"{}s{:3}ms {} {}",
  	elapsed.as_secs(),
  	elapsed.as_millis() % 1000,

Are the benefits worth a custom way of logging? Should a library instantiate a logger instead of the conventional instantiation by the calling binary?

Related, other Mozilla code using log:
https://searchfox.org/mozilla-central/search?q=%5Elog+%3D&path=*Cargo.toml&case=true&regexp=true

[larseggert]

I'd be OK with removing this.

[larseggert]

There are some special things neqo_common::log allows, e.g.,

neqo/neqo-crypto/src/agent.rs

Lines 898 to 902 in 9f0a86d

	qdebug!(
	[format!("{fd:p}")],
	"Got resumption token {}",
	hex_snip_middle(&v)
	);

Can we do this with log?

[larseggert]

So I worked on this a bit in #2291. I do think we need a wrapper of some sort like we have in neqo-common::log, because otherwise we'd need to initialize env_logger in every single test, which is a pain.

[mxinden]

Agreed. I didn't think of env_logger instantiation in unit tests. Sorry about that.

[larseggert]

OK. I will leave neqo-common::log in place w/some changes.