Refactor uses of ReloadableModule

[artem-tikhomirov]

Address uses of classloading through ReloadableModule. Now, we keep classloading aspect of a module independent of SModule class hierarchy (and independent of any repository read/write access, too).

Uses of ReloadableModule.getOwnClass() have been refactored with ModuleRuntime.getOwnClass(). ClassLoaderManager listeners (low-level, module/repository-kind of notification), have been replaced with ModuleDeploymentListener of LanguageRegistry.

[artem-tikhomirov]

@wsafonov, @sergej-koscejev the change is quite straightforwad and pretty much the same in all cases, but it touches various subsystems and I don't know what would be the right approach to review it. Split, one of you, or rather subsystem author?

[sergej-koscejev]

I don't think we want to be defensive about loaded classes not being assignable to expected classes but rather fail with an error or an exception in this case.

E.g.

Class<?> descriptorClass = module.getOwnClass(className); 
if (descriptorClass == null || !ISelectionIntentionsDescriptor.class.isAssignableFrom(descriptorClass)) { return; }
...

should be just

Class<?> descriptorClass = module.getOwnClass(className); 
if (descriptorClass == null) { return; }

[sergej-koscejev]

@artem-tikhomirov I'm going to force-update this PR myself.

[sergej-koscejev]

@artem-tikhomirov Is the result of ModuleRuntime#getOwnClass() always non-null? It says it throws CNFE, I assume it will be thrown if the class cannot be found? Could the method be annotated with @NotNull in MPS then?

[sergej-koscejev]

Changes to SelectionIntentionManager submitted as #1647.

[alexanderpann]

@sergej-koscejev can this PR be closed or merged?

[sergej-koscejev]

I’m working on adjusting individual commits and submitting them as separate PRs, please leave this open for now.