Less frequent reloads and restarts

[sed-i]

Issue

Before this change, reload/restart could result in unnecessary down time.

Solution

• Only restart if cli args changed
• Only reload if config changed
• Block if config is invalid
• Fix error + status messages
• Drive-by fixes

Fixes #342.

Context

NTA.

Testing Instructions

Afaict, the feature is fully captured in the newly added utests. Please review those for completeness!

[sed-i]

@dstathis here's a summary of StoredState vs pull for determining whether a reload is needed.

	StoredState	Pull
Multiple files	Easy to calculate hash	Need to pull (and sort?) all data (config, certs AND alerts)
Tinkering	Manual changes persist	Manual changes disappear every _configure
Robustness	Only push	Pull (multiple files) then push

Given the above I prefer StoredState. Did I miss anything?

[dstathis]

@dstathis here's a summary of StoredState vs pull for determining whether a reload is needed.

	StoredState	Pull
Multiple files	Easy to calculate hash	Need to pull (and sort?) all data (config, certs AND alerts)
Tinkering	Manual changes persist	Manual changes disappear every _configure
Robustness	Only push	Pull (multiple files) then push

Given the above I prefer StoredState. Did I miss anything?

I'm not sure I understand why we would lose local changes. I will go ahead and approve though.

[sed-i]

Patching of a bunch of ops.testing._TestingPebbleClient internals gives me bad feelings, though. Do we really need this huge amount of mock patching?

That's the least evil I could come up with at the time. Open for other suggestions.