This re-lands #41417 with a slight change that will hopefully not tickle the analyzer as it did before. The last time I tried to land this, the analyzer succeeded for the analyze step in Cirrus, and locally, but failed in an integration test.

* remove splash logic

* update tests

ReorderableListView was constructing a GlobalObjectKey using
the child key as the value. This only had the intended behavior
if the child key was identical across build method invocations.

The new strategy is to scope the child key's value to the
State object's identity, allowing child keys to have value
compare semantics while disambiguating among different list view
instances.

This reverts commit eb3e2f55 because of a suspected analyzer bug.  Reverting to investigate.

This addresses comments in the original PR (#41245) that introduced Intent.doNothing, adds tests, and fixes an issue with it.

This reverts commit e47a1dc2.

This changes the way ActionDispatchers are found by the Actions widget, so that by default it will look for dispatchers of the parent Actions widgets instead of just creating a default ActionDispatcher. This allows overriding of the ActionDispatcher at the top level: before, the custom action dispatcher would only be invoked if explicitly set on all the Actions widgets.

This is not a breaking change because there was a default value to the dispatcher parameter before that performed this function, and not specifying the dispatcher anywhere will still result in a default dispatcher being created.

The proposed change will change focus handling when pushing and popping routes so that the FocusScopeNode for the route receives focus when pushed, and that the FocusScopeNode in the navigator receives focus when the route is popped.

This means that the last setFirstFocus call on the scope is used to determine which control actually receives focus. When the focus scope receives focus, it traverses its children, trying to find a non-scope node that is the "first focus" of itself or a child node.

This is a breaking change, because the focus behavior has changed. If you push a route after this change, and had a 'first focus' set on a widget via FocusScopeNode.setFirstFocus, it won't currently receive focus immediately, but after this change it will. Similarly, if you pop a route after this change, the focus will go back to where it was before the route was pushed, which is correct, but different from what happens now.

* MouseRegion documentation claimed that onEnter and onExit
  would track entry and exit regardless of whether the pointer was
  down or up
* It did such, but when grabbing the value of `event.down` from
  the passed event, the value was always `false`
* PointerEnterEvent and PointerExitEvent were overriding the value
  passed from PointerEvent in constructors, even if the value was true
  e.g. in invocations of .fromMouseEvent((PointerMoveEvent...))
* This change now passes the value along to PointerEnter/ExitEvents
  while providing it a default of false, and updates documentation

Fixes #40637

This fixes the mouse hover code to not schedule frames with every mouse move.

Before this, it would schedule a post frame callback, and then schedule a frame immediately, even if there was nothing that needed to be updated. Now it will schedule checks for mouse position updates synchronously, unless there's a new annotation, and skip scheduling a new frame in all cases. It has to be async in the case of a new annotation (i.e. a new MouseRegion is added), since when the annotation is added, it hasn't yet painted, and it can't hit test against the new layer until after the paint, so in that case it schedules a post frame callback, but since it's already building a frame when it does that, it doesn't need to schedule a frame.

The code also used to do mouse position checks for all mice if only one mouse changed position. I fixed this part too, so that it will only check position for the mouse that changed.

* Implement system fonts system channel listener

* Carrying last size and transform information to TextInputConnection. Doing this we are making sure that the size/transform information from the previous connection will be removed, when connection changes.

* remove unused lastsize and lasttransform values

* Adding unit tests. Adressing comments.

There were four or five different implementations in various tests for sendFakeKeyEvent, which roughly all did the same thing. I was going to add yet another one, and decided that it needed to be generalized and centralized. This replaces those instances with something that just takes a LogicalKeyboardKey so that it's self-documenting, and can be used with multiple platforms.

This adds two functions to widget tester: sendKeyDownEvent and sendKeyUpEvent which simulate key up/down from a physical keyboard. It also adds global functions simulateKeyDownEvent and simulateKeyUpEvent that can be called without a widget tester. All are async functions protected by the async guard.

This reverts commit b12bdd0e as it
breaks existing tests that expect image loaded after certaing number of
pupms. With image loading done on separate isolate pumping is not
guaranteed to get image loaded.

Started reading stored channel messages from ui.channelBuffers when handlers get set.

* Fixed Selectable text align is broken

* Use separate isolate for image loading. Use TransferableTypedData to const-cost receive bytes from that isolate.

Added docs examples of getting the string from a TextField

Previously, focus nodes created in the tab scaffold were not being disposed of properly, causing possible memory leaks. Also, the builder wasn't being passed the right context so that the FocusScope.of operator inside of a builder would find the focus scope for the given tab (it was being passed the context of the overall tab scaffold).

This reverts commit 6ad88bd5.