feat: Upgrade to web-vitals@4.0.0

[wolfgangcodes]

Which problem is this PR solving?

Web vitals but released 4.0.0 which includes an updates to field names and additional for LCPAttribution, INPAttribution and TTFBAttribution and deprecated onFID. See the full changes here.

Short description of the changes

Upgrade to web-vitals@4.0.0. Full change log of the underlying changes can be found here.

For this release, we'll send both the updated field names and the deprecated field names, which will be removed in the next major release.
Additionally, onFID has been deprecated and will removed in the next major release.

For this instrumentation we have updated attribute payloads as follows:

LCP:

• Updated resource_load_time -> resource_load_duration
• Deprecated resource_load_time

INP:

• Updated element -> interaction_target
• Deprecated element
• Updated event_type -> interaction_type
• Deprecated event_type
• Added interaction_time
• Added input_delay
• Added interaction_target_element
• Added next_paint_time
• Added presentation_delay
• Added processing_duration

TTFB:

• Updated waiting_time -> waiting_duration
• Deprecated waiting_time
• Updated dns_time -> dns_duration
• Deprecated dns_time
• Updated connection_time -> connection_duration
• Deprecated connection_time
• Updated request_time -> request_duration
• Deprecated request_time
• Added cache_duration

How to verify that this has the expected result

See that the new new and updated properties are being reported.

[pkanal]

🎉 Amazing! Can we these new attributes to our docs for web vitals attributes? Up to you if you want to do it as a separate PR, that works too!

[ahrbnsn]

definitely defer to purvi & other experts here, but i'd like to vote to keep {attrPrefix}.element as the designated field (or at least dual write) — one struggle for folks less familiar with the specific data shape of various events is trying to remember what field name goes with what event type. ie instead of making people remember for inp, the element is listed under interaction_target, for cls, it's the largest_shift_target, and for lcp, it's element, keep all targets referenced under the same name — the element

more in-depth writeup/thoughts here: #61

For metrics which include reference to a specific element — largest contentful paint, cumulative layout score, interaction to next paint — that element is always included in [metric].element, instead of a metric-specific key (inp.element instead of inp.event_target, cls.element instead of cls.largest_shift_target. lcp.element remains lcp.element). Although we do lose specific signal around why the element is being referenced, I think there's benefit to folks only needing to remember one field name. Each metric has only one specific, metric-related reason for referencing an element; I don't think we're losing too much signal by shedding the specificity in the key.

[wolfgangcodes]

That's great additional context! I think that makes me more inclined to dual write. My _evolved thinking is that we should map 1-1 with the shape of web vitals data, and we can _additionally _ in a separate namespace provide more ergonomic access and/or derived data. Since we're still in EA and not removing any fields, let's merge this, and we open up a follow-up PR to tackle any changes as our thinking evolves.