core: replace EIL with MP-FID

[patrickhulce]

Summary
Promotes Max Potential FID to the spot of EIL, moves EIL to JSON only.

@patrickhulce: oh we're def making that happen for v5? 😀
@paulirish: ya

It's difficult to analyze our thresholds because the HTTPArchive data we have on max potential FID will all by impacted by the missing microtask data bug (see #8379), but MPFID differentiates *MUCH better than than EIL, they feel about right, and its still not a scored metric anyway...

Ventiles

SELECT
  APPROX_QUANTILES(CAST(JSON_EXTRACT_SCALAR(report, '$.audits.max-potential-fid.rawValue') as FLOAT64),
    20) as mpfid,
  APPROX_QUANTILES(CAST(JSON_EXTRACT_SCALAR(report, '$.audits.estimated-input-latency.rawValue') as FLOAT64),
    20) as eil
FROM
  httparchive.lighthouse.2019_03_01_mobile

Related Issues/PRs
picks up the mantle of #5842 and raises it to glory.

[brendankenny]

but MPFID differentiates *MUCH better than than EIL, they feel about right

that's because one tries to be robust to outliers and the other is "here's your outlier", but w/e :P

[brendankenny]

LGTM!

[brendankenny]

Not having the "maximum" part of the string in all the translations (until the next round comes out) isn't a huge deal, but how much do we care about getting this link out to everyone? We could manually edit the other languages to add the Learn More link.

[brendankenny]

We could manually edit the other languages to add the Learn More link.

note: this could also be a really really bad idea :)

[patrickhulce]

Ha, Given that these are hidden by default, IMO it's probably not the end of the world to update these ASAP, but others could feel differently.

[brendankenny]

Yeah, SG

[patrickhulce]

that's because one tries to be robust to outliers and the other is "here's your outlier", but w/e :P

apologies, didn't intend to be hating on EIL's usefulness as a whole. I know it's by design but it's definitively not as suitable for differentiating perf scores on top half of the web that's all