Add regex based kv extraction option

[pandemicsyn]

Changes the behaviour of the UT parser to collect any unknown kv pairs or unbalanced fields as a slice of unparsed entries (returned in the UserTraffic struct) rather than throwing an error. This is especially helpful since we somewhat frequently have duplicate value fields for keys.

For example - while auditing a sample of log lines for a Peloton site nearly 100% of the "key with no value" errors where from a country code field where the value was a duplicate like US, US.

2020/12/04 08:29:30 ==== Errors Seen ====
2020/12/04 08:29:30 key with no value: 6341 
2020/12/04 08:29:30 missing_sid_and_csid: 1926
2020/12/04 08:29:30 lines seen: 19058
2020/12/04 08:29:30 bytes potentially billed: 611126576
2020/12/04 08:29:30 bytes potentially missed: 45379903

[rybit]

I think that this is an ok solution, I'm wondering if it might be easier to actually not error on the case of a key with no value. We could just log a warning (or a counter?) and then move on?

[pandemicsyn]

Yea - we have that catch all "other" map as the default case for when we run into unknown kv pair's:

netlify-commons/parser/user_traffic.go

Lines 154 to 158 in d1232d6

	default:
	if ut.Other == nil {
	ut.Other = make(map[string]string)
	}
	ut.Other[parts[0]] = parts[1]

I tested a variation that just adds the offending "key with no value" error entries as key's with an empty value to that others map. Then folks using the parser can check that map for entries and generate log lines/metrics accordingly. I'd prefer that to adding logging/metrics to the parser directly. I'm definitely down for doing that instead of the regex. @rybit wdyt ?

[rybit]

yeah I think that approach works nicely. Or maybe we change the 'other' to be a list of unparsable sections, the whole bit.