Refactor time_restriction and stabilise plans containing notification_rule.steps, notification_policy.filter.conditions and time_restrictions

[Baarsgaard]

Commits:

1. Closes Docs for "opsgenie_notification_policy" contain invalid syntax #246

2. Closes opsgenie_notification_rule.steps is order sensitive #253 And opsgenie_notification_policy conditions order is not stable causing "in-place" updates though nothing changes #409
   And adds some much needed schema validation for Terraform-ls to use.

3. Closes Fix for "time restriction" is not stable #341:
   Unifies the Schema, Input-Validation, and expanding/flattening for time_restriction blocks.
   This implementation should fix unstable plans containing time_restriction for all resources.
   Schedule rotation "time restriction" is not stable #225 only addressed a single resource.

4. Closes opsgenie_notification_policy not validating that all the expected fields are defined #355
   Way stricter input validation of actions before terraform apply

This is a pretty big PR solely because of the time_restriction refactor, I am willing to split it, but I do still think reviewing this PR one commit at a time is feasible.

[Baarsgaard]

@koushik-swaminathan Sorry to ping you, but I would love to have this reviewed to continue using the provider :)

[koushik-swaminathan]

@Baarsgaard sure, I'll take a look. Have you tested these changes?

[Baarsgaard]

Yes, all of it is tested.
Luckily very little of the actual business logic has changed.
The only new code is the resourceOpsGenieNotificationPolicyMultiValueValidation() function in the 3rd commit, the rest is refactoring and Schema validation updates.

[koushik-swaminathan]

LGTM, tested it as well. But it looks like it doesn't really fix the conditions in-place issues, but time-restriction seems to be stable now.

[Baarsgaard]

Could you share what specifically does not work?
I've been testing with a fair amount of notification policies and I'm unable to get the perpetual state drift to show up:

Terminal cast

You can see me changing the order of every single condition, both using message and tags
This causes perpetual drift on the current release.

[koushik-swaminathan]

Hi @Baarsgaard, the in-place issue was persisting in my local but it seems to work for you. LGTM, approving the PR.

[koushik-swaminathan]

Could you edit the date and version in the changelog? I'll merge the PR after that. Thank you!

[Baarsgaard]

I'll set it for Monday next week 2023-11-06

[Baarsgaard]

@koushik-swaminathan Should be ready for a merge now! :D

[Baarsgaard]

LGTM, tested it as well. But it looks like it doesn't really fix the conditions in-place issues, but time-restriction seems to be stable now.

I was thinking about what could be the cause of this.
The difference is likely that the policies I created were saved as a set, but your test was made on policies stored as the old list.

If you run terraform apply -refresh-only followed by a normal terraform apply is the plan still unstable?

[koushik-swaminathan]

LGTM!