fix!: have insertion markers use json deserialization

[BeksOmega]

The basics

• I validated my changes

The details

Resolves

Fixes #7316

Proposed Changes

Reverts #7430 #7429

Makes it so that insertion markers are created using JSON deserialization.

Reason for Changes

From #7364

Insertion markers duplicate the block being dragged, and then tell the duplicate to be an insertion marker.

Currently the duplication duplicates (ha) a lot of code from the serialization system. We think this is because in the past the XML system didn't have a good way to just serialize and deserialize a single block (without children). But the JSON system does have this!

Sparked by this discussion: https://groups.google.com/g/blockly/c/gcvynl88QLY/m/shFyX4x0BwAJ

Test Coverage

All tests pass.

Documentation

Additional Information

N/A

Breaking changes / To fix

Insertion-marker-ness is now set after any mutation is deserialized.

If you have a block whose behavior depeonds on whether it is an insertion marker or not, they should check doFullSerialization instead. Insertion markers pass doFullSerialization as false. Ensure that if your block has a backing data model, insertion markers either properly create and destroy those models, or properly reference and do not destroy those models.

You can see the shareable procedure blocks for an example of round-tripping whether doFullSerialization was enabled from saveExtraState to loadExtraState.

[maribethb]

LGTM, but please fix the description of the PR to include the directions for breaking changes directly instead of describing what would be in the release notes