fix: fluent interface kernel session

[mwdchang]

Summary

This PR handles two things:

• Instead of returning null if kernel-session-manager's sendMessage fails, throw an error. This makes the fluent interface a bit less gnarly to read. Instead of sendMessage(...)?.register, now it is just sendMessage(...).register

• I noticed we send a lot of extra, unneeded messages (e.g. reset) for model-edit, this cleans them up

Testing

No functional changes, can test with one or more operators that leverages the kernel session manager.

For model-edit:

• pipe in a model to model-edit operator,
• in the notebook we can override with an empty mira-model-template

# Define an empty model as basis
model = TemplateModel(
	templates = [],
	parameters = {},
	initials = {},
    observables = {},
	time = Time(name = "t", units = None),
	annotations = Annotations(name = "PetriNet model created by a user"))

For code-to-mode:

• create a code-to-model operator
• select julia, select deapodes
• run the following code

  C::Form0{X}
  dC::Form1{X}

  dC == d₀{X}(C)