Update types to support more generic usage

[brendankenny]

Fixes #408

Relatively light type clean up for v4. Things are looking pretty good and I ended up not changing that much, which is great. For many use cases there won't be breaking changes here in practice (e.g. if you already had a LCPMetric specifically, the name of the base interface or a union it might belong to doesn't matter).

It's possible some things still went too far for some sensibilities; let me know and I can just roll those back. Once this is looking good, I'll update the README entries around these.

Changes:

• Switch MetricType for Metric (as originally in #359)

  Most of the time as a user you just want to know which of the union of possible metric types you have, you don't want the combined type in the old Metric. If you do somehow end up with a generic Metric, it's now a discriminated union, so checking e.g. metric.name === 'LCP' will narrow the type to LCPMetric. The old Metric is still available as MetricBase (open for name bike shedding)

• Ditch CLSReportCallback/CLSReportCallbackWithAttribution, etc. Maybe controversial, but IMO these end up as too many named types for things users will usually never explicitly type, and the only important part is the parameter type, so let's just use that directly.

• slight clean up of unattributedOn* methods

  e.g. unattributedOnCLS was supposedly calling back with a metric of type CLSMetricWithAttribution, which isn't true. metric is a CLSMetric and only becomes a CLSMetricWithAttribution after attributeCLS is called. Still a type assertion as CLSMetricWithAttribution in there, but it better reflects what's really going on. We could have the attribute* methods return the newly attributed object to avoid the extra type assertion, but I wasn't sure if that was too big a change.

• A few scattered removed unnecessary type assertions, e.g. as PerformanceEventTiming when the compiler already knows they're PerformanceEventTimings

[brendankenny]

We could have the attribute* methods return the newly attributed object to avoid the extra type assertion, but I wasn't sure if that was too big a change.

I like this way better, actually, so I added it. Separate commit so it's easy to remove, though.

[brendankenny]

The main thing missing right now is to copy the type changes to the relevant parts of the README documentation

PTAL

[philipwalton]

@brendankenny I made some updates to your updates (and merged in @tunetheweb's test fixes). Let me know if you have any concerns with any of the changes I made.

[tunetheweb]

Oh and just realised there's a merge conflict due to #468 being merged.

[philipwalton]

Actually I ended up just removing all the links to the source types because it looks like they no longer work with GitHub's new file viewer UI.