Handle legacy crash logs and use proper architecture in symbolicate.js

[ayoy]

Task/Issue URL: https://app.asana.com/0/0/1202475379856923/f

Description:
I updated symbolicate.js to:

• specify crash log architecture while looking up symbols in the dSYM file
• handle old crash log format in addition to IPS
• fix symbolicating IPS crash logs that contain Application-Specific Backtrace (previously the script wouldn't symbolicate symbols outside of asiBacktraces)
• recognize IPS crash format based on file content, not extension (to handle crashlogs downloaded from Asana more seamlessly) – after symbolication, an IPS file is renamed from .crash if needed
• the script does not require the app binary to be present anymore (it only uses the dSYM file)

Steps to test this PR:

1. Download some valid crash reports from Asana (i.e. displaying correct app version), put them in the directory, e.g. scripts/symbolicate/crashes.
2. Download dSYM files from Asana and put them in scripts/symbolicate/binaries according to scripts/symbolicate/binaries/README.md (see screenshot).
3. Change directory to scripts/symbolicate and run the script: ./symbolicate.js crashes.
4. Verify that crash logs are symbolicated and IPS files have .ips extension.

---

Internal references:

Software Engineering Expectations
Technical Design Template
When ready for review, remember to post the PR in MM

[samsymons]

LGTM, this works great when you've got a complete crash report.

I initially tried with a crash report where the app was killed due to Code Signature Invalid, and for that report the list of images used was incomplete and didn't have the correct names. The script of course failed in that case - but I don't know what else we can do with crashes like that.

If we were to go further down this road, I would suggest using Git LFS so that we can store dSYM files there and make it easier on us to pull these down automatically – but since we have the chance of moving to automatic symbolication using our own remote instance, I don't think it's worth thinking about for now.

[ayoy]

Thanks @samsymons! That's right, the error handling could be improved, but as it's only a semi-automatic tool for now (needs crashlogs to be downloaded manually before symbolication) I'm solving it by just not downloading "broken" crash reports :) Agreed that Git LFS would do the job (or even uploading dsyms to GitHub releases!). I'm still mentally in the closed-source software world and I haven't thought about it, but in fact our debug symbols will eventually not be a secret :)