Split E2E runs into batches and randomize order

[nickdowell]

Goal

We have observed test flakes that appear to be Appium related occur around the 1hr mark during an E2E test run.

Changeset

Split all E2E runs into batches to reduce the duration of each Appium / BS session. Note that batch 1 takes longer than batch 2 because there are a large number of scenarios in files beginning with e that prevent an even split.

Enabled --order=random to make it easier to differentiate between flakey scenarios and time-related issues.

Made the step naming more consistent and concise.

Testing

For a sample full run see https://buildkite.com/bugsnag/bugsnag-cocoa/builds/2709

[github-actions]

Infer: No issues found 🎉

OCLint: No issues found 🎉

Bugsnag.framework binary size did not change - 1,126,840 bytes

Generated by 🚫 Danger

[nickdowell]

there are a large number of scenarios in files beginning with e that prevent an even split.

I don't understand - the split is on a-n and o-z. Can you just push m and n into batch 2? (I wouldn't get too hung up, though, I found they can be variable).

The split has been changed to to a-g and h-z but batch 1 is still the heaviest.

If we push e into batch 2 then we'll just reverse the problem...

[twometresteve]

LGTM - just one comment on the pipeline comment.