Revert Chrome details for PerformanceResourceTiming.{finalResponseHeadersStart,responseStart}

[tunetheweb]

Summary

This reverts most of #25733

The aim of that PR (as well as adding finalResponseHeadersStart) was to try to explain that Chrome diverged in it's definition of responseStart for sites using Early Hints. However, in hindsight, I think the updates are causing more confusion than help.

MDN currently shows this:

This suggests the header was only added in 133 (it wasn't it was supported since 85 with a slight change in definition for some sites).

It also suggests it's not supported yet in Opera, Opera Android, and Samsung internet when that's not the case.

All in all, I think my intent to help developers has caused more issues than it solved, and we should instead treat this as a small oddity in Chrome 115-132 that doesn't require documenting in BCD (like some bugs aren't unless major issues).

Additionally, I incorrectly noted this was available from Chrome 87, when it was actually available from Chrome 43 so that was a typo.

It's also causing confusion in downstream dependencies:

microsoft/TypeScript-DOM-lib-generator#1891

Similarly I think the noting of finalResponseHeadersStart as being available earlier, with the alternative name of responseStart is more confusing than helpful so also removing that.

Test results and supporting details

N/A - see #25733

Related issues

#25733

[caugner]

This suggests the header was only added in 133 (it wasn't it was supported since 85 with a slight change in definition for some sites).

This might be a UX problem, because we always show only the version were full support was added, and this may be misinterpreted as "it wasn't added before".

All in all, I think my intent to help developers has caused more issues than it solved, and we should instead treat this as a small oddity in Chrome 115-132 that doesn't require documenting in BCD (like some bugs aren't unless major issues).

The alternative would be to add a note like:

Before Chrome 133, the property reflected the time of the final response, even if an interim response is available.

I think it is reasonable in this case.

/cc @Elchi3 for a second opinion.

[tunetheweb]

Yeah Caniuse interprets the data more as I hoped MDN would:

Also the note is only relevant from 115 to 132 so not "before 133". Given this is a relatively short period from 43 - current, not sure a note for the full time is the best UI either.

Also it's only sites using Early Hints, and where they have a large gap between the Early Hint and the final response, that are affected. So it's a bit niche to be honest.

Finally Chrome actually did the right thing here (we updated the spec as that's what was agreed to do), but other browsers didn't follow so we've decided to go for another non-breaking change (and to revert the spec back).

All in all I think just reverting it is the best.

[caugner]

Sounds plausible.

If @Elchi3 has no concerns, let's go with the revert.