context

problem:
- we are starting to get captcha required errors on prod
- this is bad b/c it blocks us from: (1) registering new phone numbers, (2) re-registering old phone numbers to recover from a corrupted keystore
solution:
- introduce a -c argument to boost register that accepts a captcha token, which may be obtained from https://signalcaptchas.org/registration/generate.html
- introduce an optional captchaToken field into the /phoneNumbers/register endpoint hit by boost register -- and pass it down the callstack to signald.register, as per: https://gitlab.com/thefinn93/signald/-/wikis/Captchas

changes

support captchaToken argument in register endpoint

add captcha token arg (-c) to boost register

modify signal configs

decrease registration batch size from 5 to 4 (to avoid triggering rate limit errors on excessive registrations)
increase signaldRequestTimeout from 10 to 20 (since this is now effectively the timeout for verifications)

fix _check-env bug causing non-deterministic ngrok discovery

symptom: sometimes _check-env incorrectly deletes the first character from the ngrok tunnel url assigned to SIGNALBOOST_HOST_URL in dev mode, with the result that twilio numbers are created with incorrect callback urls and we can never receive sms'es sent to this number and thus never successfully verify new numbers with signal
cause: we wrongly assumed that ngroks tunnels endpoint would always return the http and https tunnels in the same order and based our sub-string finding command off of this assumption -- stripping the first 8 digits from the public_url field in the first tunnel on the assumption that this tunnel's url would always start with https://. however, sometimes it starts with http://, and in those cases, we trim the first digit from the tunnel url
fix: use sed to deterministically strip either http:// or https:// from the tunnel url so that the ordering doesn't matter

Edited Aug 10, 2020 by aguestuser