Lighthouse: Sample our sentry errors more heavily

[paulirish]

we have a lot of high volume errors that are just happening too often.

Here's our top 25 errors in the last 24hr:

"Failed to read the 'cssRules' property from 'CSSStyleSheet': Cannot access rules": 305k,
"Could not find any top level events": 249k,
"No firstMeaningfulPaint found, usingfallback: 214k,
"Invalid task timing data": 202k,
"The asynchronous expression exceeded the allotted time of 60s": 18k,
"Could not load stylesheet (timeout)": 104k,
"Invalid Version: <version>": 103k,
"Invalid image sizing information <url>": 76k,
"Cannot create DNS cache with no rtt": 70k,
"finalUrl <url> not found in network records.": 50k,
"SPEEDINDEX_OF_ZERO": 34k,
"NO_TTI_NETWORK_IDLE_PERIOD": 18k,
"NO_FMP": 33k,
"Images have different sizes!": 29k,
"Cannot simulate graph with cycle": 28k,
"Cannot read property 'indexOf' of undefined": 24k,
"Cannot read property 'after' of undefined": 16k,
"Failed to construct 'URL': Invalid URL": 14k,
"navigationStart was not found in the trace": 14k,
"Failed to execute 'matches' on 'Element': '<selector>' is not a valid selector.": 13k,
"Failed to execute 'getComputedStyle' on 'Window': parameter 1 is not of type 'Element'.": 11k,
"NO_FCPUI_IDLE_PERIOD": "106k,
"NO_FCP": "102k

in sentry.js you'll see we sample some of these. we could afford to sample more... well actually, it probably makes sense to sample everything. lets start at 1%?

[brendankenny]

we should throttle it in Sentry.init and just not init anything if in the 99% (just return like if reporting was disabled) so breadcrumbs and stuff don't get messed up

[SamHasler]

For anyone else googling "Cannot create DNS cache with no rtt" you can fix that by going to chrome://flags/ and disabling "Simple Cache for HTTP"