Trying to speed up type handling

[oxisto]

This PR should speed up type handling immensely. The list of first oder types is now a map of a list instead of a list. The issue is that we have a lot of "types" (remember, a "type" is a unique combination of a scope and a name), we potentially have many scopes and a simple "myType" can exist as multiple Type nodes before the type resolver harmonises this. In order to at least have unique types per scope, we iterate existing types and return one that matches the name/scope combination. This was previously done over the complete list and is now only restricted to those types whos name matches in the first place.

This is more a workaround than a complete solution to the problem of types but should at least be significantly faster now.

As a benchmark I parsed http://github.com/openstack/glance. This speeds up the parsing from 73s to 42s.

Before:

Metric	Value
TranslationManager: Executing Language Frontend	16422 ms
CompressLLVMPass: Executing Pass	2 ms
ImportResolver: Executing Pass	528 ms
TypeResolver: Executing Pass	988 ms
TypeHierarchyResolver: Executing Pass	425 ms
CXXExtraPass: Executing Pass	215 ms
GoExtraPass: Executing Pass	327 ms
PythonAddDeclarationsPass: Executing Pass	1278 ms
ResolveCallExpressionAmbiguityPass: Executing Pass	4378 ms
JavaExternalTypeHierarchyResolver: Executing Pass	2 ms
EvaluationOrderGraphPass: Executing Pass	698 ms
JavaImportResolver: Executing Pass	0 ms
SymbolResolver: Executing Pass	33495 ms
ControlDependenceGraphPass: Executing Pass	6523 ms
DFGPass: Executing Pass	792 ms
ControlFlowSensitiveDFGPass: Executing Pass	3183 ms
DynamicInvokeResolver: Executing Pass	433 ms
ProgramDependenceGraphPass: Executing Pass	2295 ms
PrepareSerialization: Executing Pass	807 ms
FilenameMapper: Executing Pass	320 ms
TranslationManager: Translation into full graph	73128 ms

After:

Metric	Value
TranslationManager: Executing Language Frontend	15854 ms
CompressLLVMPass: Executing Pass	6 ms
ImportResolver: Executing Pass	575 ms
TypeResolver: Executing Pass	856 ms
TypeHierarchyResolver: Executing Pass	198 ms
CXXExtraPass: Executing Pass	201 ms
GoExtraPass: Executing Pass	311 ms
PythonAddDeclarationsPass: Executing Pass	1360 ms
ResolveCallExpressionAmbiguityPass: Executing Pass	274 ms
JavaExternalTypeHierarchyResolver: Executing Pass	2 ms
EvaluationOrderGraphPass: Executing Pass	835 ms
JavaImportResolver: Executing Pass	1 ms
SymbolResolver: Executing Pass	7557 ms
ControlDependenceGraphPass: Executing Pass	7046 ms
DFGPass: Executing Pass	807 ms
ControlFlowSensitiveDFGPass: Executing Pass	3399 ms
DynamicInvokeResolver: Executing Pass	311 ms
ProgramDependenceGraphPass: Executing Pass	1667 ms
PrepareSerialization: Executing Pass	626 ms
FilenameMapper: Executing Pass	253 ms
TranslationManager: Translation into full graph	42154 ms

[maximiliankaul]

Speedup (start to end cpg parsing (frontend + passes)) against current main is about 4x with a larger Python test case 👍

[oxisto]

Unfortunately, while this speeds up python, this slows down the CXXExtraPass extremely :(

[sonarcloud]

Please retry analysis of this Pull-Request directly on SonarCloud