* tshd events: Wait for listeners before responding

When tshd is sending the relogin event, it needs to be able to know when
the Electron app has finished relogging the user. I wanted to implement
this by simply waiting for the response from the RPC.

I'm so glad we did not use gRPC streams for tshd events as this would be
much harder to implement with streams.

* Return a function from ModalsService.openDialog which closes dialog

With the introduction of important and regular modals, this will help us
close the specific dialog if need arises.

* Make tshd events listeners aware of request cancellation

This will be useful in the upcoming commits. Basically, tshd is going to
ask the Electron app to relogin the user, with a 1 minute timeout.
The Electron app will show a login modal but if the user doesn't submit
it within 1 minute, tshd is going to cancel the request.

In that situation, we need to be able to close the modal.

* Add support for passing reason in DialogClusterConnect

* Add support for important modals

This will let us show the relogin modal on expired cert, even if the user
was using some other modal at that moment.

* Remove title attr from notification text

The user can read more by expanding the notification. The title attribute
persisted even after expanding the notification, making reading it harder.

* Add WindowsManager.forceFocusWindow

* Use IAppContext instead of AppContext

The next commit is going to add a private method to AppContext.
IAppContext is an interface which enables us to pass a mocked version of
AppContext in tests. That mock is not going have that private method,
so any place accepting AppContext wouldn't be able to accept the mocked
AppContext.

Instead, classes & functions should accept IAppContext rather than AppContext.

* Implement handlers for new tshd events

tshd needs to be able to do two things:

- Ask the user to log in again.
- Forward errors from goroutines running gateways to the Electron app
  in form of a notification. Otherwise those error would be visible
  only in the logs.

* Don't restart gateways after logging in

Restarting the gateways on login was a workaround from times where gateways
didn't manage their own certs.

In the new flow, a gateway takes care of refreshing the certs itself
through the middleware passed to alpnproxy.LocalProxy.

syncRootClusterAndRestartClusterGatewaysAndCatchErrors used to call two
functions:

- syncRootClusterAndCatchErrors
- restartClusterGatewaysAndCatchErrors

The second function is no longer necessary, so we can make any place that
was calling syncRootClusterAndRestartClusterGatewaysAndCatchErrors
call just syncRootClusterAndCatchErrors instead.

…1416)

* tshd events: Wait for listeners before responding

When tshd is sending the relogin event, it needs to be able to know when
the Electron app has finished relogging the user. I wanted to implement
this by simply waiting for the response from the RPC.

I'm so glad we did not use gRPC streams for tshd events as this would be
much harder to implement with streams.

* Return a function from ModalsService.openDialog which closes dialog

With the introduction of important and regular modals, this will help us
close the specific dialog if need arises.

* Make tshd events listeners aware of request cancellation

This will be useful in the upcoming commits. Basically, tshd is going to
ask the Electron app to relogin the user, with a 1 minute timeout.
The Electron app will show a login modal but if the user doesn't submit
it within 1 minute, tshd is going to cancel the request.

In that situation, we need to be able to close the modal.

* Add support for passing reason in DialogClusterConnect

* Add support for important modals

This will let us show the relogin modal on expired cert, even if the user
was using some other modal at that moment.

* Remove title attr from notification text

The user can read more by expanding the notification. The title attribute
persisted even after expanding the notification, making reading it harder.

* Add WindowsManager.forceFocusWindow

* Use IAppContext instead of AppContext

The next commit is going to add a private method to AppContext.
IAppContext is an interface which enables us to pass a mocked version of
AppContext in tests. That mock is not going have that private method,
so any place accepting AppContext wouldn't be able to accept the mocked
AppContext.

Instead, classes & functions should accept IAppContext rather than AppContext.

* Implement handlers for new tshd events

tshd needs to be able to do two things:

- Ask the user to log in again.
- Forward errors from goroutines running gateways to the Electron app
  in form of a notification. Otherwise those error would be visible
  only in the logs.

* Don't restart gateways after logging in

Restarting the gateways on login was a workaround from times where gateways
didn't manage their own certs.

In the new flow, a gateway takes care of refreshing the certs itself
through the middleware passed to alpnproxy.LocalProxy.

syncRootClusterAndRestartClusterGatewaysAndCatchErrors used to call two
functions:

- syncRootClusterAndCatchErrors
- restartClusterGatewaysAndCatchErrors

The second function is no longer necessary, so we can make any place that
was calling syncRootClusterAndRestartClusterGatewaysAndCatchErrors
call just syncRootClusterAndCatchErrors instead.