Support for hapi v21, node v18, ESM tests

[devinivy]

• Support and test on hapi v21, drop support for hapi v19.
• Support and test on node v18, utilize some newer JS syntax, switch to hrtime.bigint() rather than legacy hrtime().
• Update all deps for node v18 and hapi v21.
• Test ESM support.

If this looks good, this will go out as bounce v10 (including #128).

[devinivy]

@Marsup process.hrtime() has been marked as "legacy", and node encourages folks to switch to process.hrtime.bigint(). It's also a bit easier to use since BigInt is a more standard type than the tuple that process.hrtime() returns.

[Marsup]

My bad, it was even written in the description, sorry 🙁 Good to merge 👍🏻

[kanongil]

@Marsup process.hrtime() has been marked as "legacy", and node encourages folks to switch to process.hrtime.bigint(). It's also a bit easier to use since BigInt is a more standard type than the tuple that process.hrtime() returns.

performance.now() would probably be an even better replacement. It's web-compatible and avoids the use of BigInt.

[Marsup]

h2o2 is very unlikely to be used anywhere else though.

[kanongil]

For sure, but if you are updating it, you might as well use that. hrtime() should only be used if you need to accurately know the current time (ie. not only used to compute a delta).

[devinivy]

I thought that too until recently—I think it's a common misconception! Here's what the docs say about hrtime:

These times are relative to an arbitrary time in the past, and not related to the time of day and therefore not subject to clock drift. The primary use is for measuring performance between intervals

[kanongil]

Oh, right! – it is not even useful for getting the exact time. Guess it is just completely redundant then.

FYI, the node performance.now() implementation uses hrtime(), but uses the start of of the current process as a reference, rather than the start of the system.