timers: recent performance regressions

[mscdex]

0d9f3bd introduced a noticeable performance regression with various timers benchmarks. For example:

• timers/immediate.js type="clear" performance dropped 77%
• timers/timers-cancel-pooled.js performance dropped 55%
• timers/timers-timeout-pooled.js performance dropped 41%

and many other timers benchmarks/benchmark configs saw double digit drops.

While having a public "get active requests"/"get active handles" API would be nice, I don't think it is worth this kind of penalty if we can't find some way to remedy the performance regressions.

/cc @RaisinTen

[RaisinTen]

To mitigate this, we can traverse

node/lib/internal/timers.js

Line 161 in c8f5dd6

const timerListMap = ObjectCreate(null);

for the Timeouts and

node/lib/internal/timers.js

Line 426 in c8f5dd6

const outstandingQueue = new ImmediateList();

for the Immediates. That way, we wouldn't have to run any extra code when the timers are added/removed. The results are going to be a little less accurate because these objects aren't updated at the right moment always (which is why I didn't do it that way in #40813) but we can always fix that in a separate PR.

[RaisinTen]

PR: #41231

[Trott]

we can always fix that in a separate PR

And if we can't fix it, we can always just leave it documented as an API limitation.