settings: allow overriding the default for settings

[andreimatei]

Release note: None

[cockroach-teamcity]

This change is 

[darinpp]

Looks fine. I do prefer slightly the simpler solution (that is currently in the branch) that doesn't require changing each of the setting classes.
Don't you need to change all setting classes in your case? StringSetting and StateMachineSetting as well?

[andreimatei]

There's two reason why I've done it this way:

1. Ergonomics. I think it's much nicer to change the default for a setting only by calling methods on that setting (instead of also having to interact with a SV)
2. Consistency. Dealing with defaults values is already the responsibility of the different Setting implementations (the setToDefault() methods).

Don't you need to change all setting classes in your case? StringSetting and StateMachineSetting as well?

I don't need to. The Override method is not part of the Setting interface, and nor should it be. For StateMachineSetting in particular, there's no notion of a "default", so there should also not be a notion of overriding it. That's another plus if this patch vs the alternative.

I'll merge this also because it's more ready then the alternative - it's its own commit (which it has to be) and it has a test :p

bors r+

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @RaduBerinde)

[craig]

Build failed

• GitHub CI (Cockroach)

[andreimatei]

failure because of broken/polluting existing test. Fixed in a separate new commit.

bors r+

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @RaduBerinde)

[craig]

Build failed

• GitHub CI (Cockroach)

[andreimatei]

bors r+

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @RaduBerinde)

[craig]

Build failed

• GitHub CI (Cockroach)

[andreimatei]

failure is some TCL flake, although I can't say what exactly

bors r+

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @RaduBerinde)

[craig]

Build succeeded

• GitHub CI (Cockroach)