* make compiler worker count configurable

go/flutter-pipeline-improvements for more details.

Classs for binary message encoding/decoding that matches the ones in the Flutter services library.

This roll unforks the Dart SDK to use the non nullable SDk,
the engine platform file is built with all the dart:* libraries
opted in except for dart:ui

dart-lang/sdk@617bc54b71 [dart2js] unit test patch for unfork cl.
dart-lang/sdk@ce5d18b0f0 [CFE] - Update exectations of CFE unit tests to account for NNBD unfork
dart-lang/sdk@be0c78a7cc [dart2js] Fix platform-binaries on bots.
dart-lang/sdk@03ce33f80c [BUILD] - Initial CL to unfork the NNBD Dart SDK

FlValue is a lightweight object used to contain the value types that Flutter
uses on platform channels.

* Implement linear gradient matrix
* update golden locks
* Remove matrix check in RadialGradient

This reverts commit a7a25d3b and relands our reland https://github.com/flutter/engine/pull/17915.

Additionally, we fixed the cull rect logic in `OpacityLayer::Preroll` which is  the root cause of https://github.com/flutter/flutter/issues/56298. We've always had that root problem before but it did not trigger performance issues because we were using the OpacityLayer's `paint_bounds`, instead of its child's `paint_bounds` for preparing the layer raster cache. A correct handling of the cull rect should allow us to cull at any level.

It also turns out that our ios32 (iPhone4s) performacne can regress a lot
without snapping. My theory is that although the picture has a
fractional top left corner, many drawing operations inside the picture
have integral coordinations. In older hardwares, keeping those
coordinates integral seems to be performance critical.

To avoid flutter/flutter#41654, the snapping
will still be disabled if the matrix has non-scale-translation
transformations.

This does some long-overdue refactoring of the spaghetti code that grew in the GLFW embedding to begin providing a clearer separation between the engine and the window. It is now possible to register plugins, and run the runloop, on a headless engine, which makes headless mode much more usable. This is useful in some automated testing environments.

There is more refactoring that should be done in the future, but this is a good incremental point to stop as the PR is already large, and it provides useful new functionality as-is.

dart-lang/sdk@39e0e75fcf Remove ElementSuggestionBuilder
dart-lang/sdk@2707880f1b [vm] Fix various UBSan failures.
dart-lang/sdk@8c79102889 [vm/isolates] Move saved_initial_field_table to isolate group.
dart-lang/sdk@9eff98594f Cleanup unnecessary references to experiments that are enabled by default
dart-lang/sdk@d4514bef78 Bypass pub constraint validation for dartfix 0.1.7.
dart-lang/sdk@30da9d580b Ensure new analysis drivers are seeded with existing priority files
dart-lang/sdk@f16663a678 Convert LocalReferenceContributor to use SuggestionBuilder
dart-lang/sdk@7cbd5f6f78 [cfe] Add round-trip serialization support for loop nodes
dart-lang/sdk@9491c6cc48 [dart2js] Fix codegen/variance_subtype_cast_test

Fixes https://github.com/flutter/flutter/issues/49398

* Handle leak of message handle when no engine present
* Move callback in wrong location

The engine was using a global to store a timestamp representing the
launch of the engine.  This timestamp is initialized with a JNI call
on Android and during shell setup on other platforms.  Later the
timestamp is added to a FlutterEngineMainEnter timeline event used to
measure engine startup time in benchmarks.

This PR removes the global and the JNI call and moves the timestamp
into the settings object.