Core: Fix operator[] for typed dictionaries

[Repiteo]

• Fixes Typed Dictionaries can allow different typed keys/values when using operator[] #96772

Caused by an erroneous regression when rebasing the typed dictionary PR. Reimplements type checking and passing _p->typed_fallback to erroneous keys when using operator[] on dictionaries.

This is technically a partial fix, because it seems the untyped value assignment via operator[] is something that can happen regardless. However, fixing that will probably be a much more involved change, as the passed value isn't linked to the typed dictionary in isolation.

[dalexeev]

Looks good to me, but it would be nice to get a second review.

[dalexeev]

(Just a note) Perhaps we should stick to one style for comment keywords.

Occurrences

core/io/packet_peer.cpp: 1
302:  2: 	//warning may lose packets
core/math/transform_2d.h: 2
 42:  2: 	// Warning #1: basis of Transform2D is stored differently from Basis. In terms of columns array, the basis matrix looks like "on paper":
 50:  2: 	// Warning #2: 2D be aware that unlike 3D code, 2D code uses a left-handed coordinate system: Y-axis points down,
editor/plugins/tiles/tile_set_atlas_source_editor.h: 1
180:  3: 		// Warning: keep in this order.
modules/gdscript/gdscript_compiler.cpp: 1
2634:  1: // Warning: this function cannot initiate compilation of other classes, or it will result in cyclic dependency issues.
modules/gdscript/gdscript_parser.cpp: 1
4319:  3: 		// WARNING: Do not merge with the previous `if` because there `!=`, not `==`!
modules/mono/glue/GodotSharp/GodotSharp/Core/GD.cs: 1
 29:  9:         /// Warning: Deserialized object can contain code which gets executed. Do not use this
platform/windows/display_server_windows.cpp: 1
3965:  2: 	// WARNING: we get called with events before the window is registered in our collection
platform/windows/gl_manager_windows_native.cpp: 1
455:  2: 	// WARNING: p_window_id is an eternally growing integer since popup windows keep coming and going
servers/physics_2d/godot_body_pair_2d.cpp: 1
165:  1: // Warning: the way velocity is adjusted down to cause a collision means the momentum will be weaker than it should for a bounce!
servers/physics_2d/godot_step_2d.cpp: 2
254:  2: 	// Warning: This doesn't run on threads, because it involves thread-unsafe processing.
261:  2: 	// Warning: _solve_island modifies the constraint islands for optimization purpose,
servers/physics_3d/godot_body_pair_3d.cpp: 1
165:  1: // Warning: the way velocity is adjusted down to cause a collision means the momentum will be weaker than it should for a bounce!
servers/physics_3d/godot_step_3d.cpp: 2
358:  2: 	// Warning: This doesn't run on threads, because it involves thread-unsafe processing.
365:  2: 	// Warning: _solve_island modifies the constraint islands for optimization purpose,
servers/rendering/renderer_rd/shaders/effects/ssao.glsl: 2
 53: 54: #define SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET (2) // !!warning!! the MIP generation on the C++ side will be enabled on quality preset 2 regardless of this value, so if changing here, change
 56:  1: // !!warning!! the edge handling is hard-coded to 'disabled' on quality level 0, and enabled above, on the C++ side; while toggling it here will work for
servers/rendering/renderer_rd/shaders/effects/ssil.glsl: 1
 52:  1: // !!warning!! the edge handling is hard-coded to 'disabled' on quality level 0, and enabled above, on the C++ side; while toggling it here will work for

[Repiteo]

I'm personally a big fan of the "proper" syntax implemented in gdextension_interface.h, though that only works in a docstrings & is primarily suited for exposed/binding contexts. I'm down for a universal WARNING: preface, akin to TODO: & FIXME: , though that'll need to be specified in the documentation.

[akien-mga]

Looks good to me.

[akien-mga]

Thanks!