Previously ballerina `readonly` return type allowed for java void methods. However, `error?` was not allowed with java void. This inconsistency is due to type checking not recognizing the fact that readonly type does include the error type. The error given in the latter case too is wrong because error is optional.

Previously this was implemented with a temporary workaround which was inefficient as we recalculated the semtypes.

Please refer to #43344 (comment) for the reason

 Rewrite BUnionType's `getMemberTypes()` usages using SemTypes

# Conflicts:
#	.github/workflows/pull_request_full_build.yml
#	.github/workflows/pull_request_ubuntu_build.yml
#	.github/workflows/pull_request_windows_build.yml
#	compiler/ballerina-lang/src/main/java/org/wso2/ballerinalang/compiler/bir/codegen/JvmTerminatorGen.java
#	compiler/ballerina-lang/src/main/java/org/wso2/ballerinalang/compiler/bir/codegen/methodgen/ModuleStopMethodGen.java
#	compiler/ballerina-lang/src/main/java/org/wso2/ballerinalang/compiler/bir/codegen/split/values/JvmObjectGen.java
#	compiler/ballerina-lang/src/main/java/org/wso2/ballerinalang/compiler/desugar/Desugar.java

Sync nutcracker branch for java 21 changes

Avoid repeated type creation

Cache int singleton creation

…_resolve

Refactor semtype creation

Use new api in CompilerTypeTestApi

Extract out the which semtype implementation used for tests

Implement semtype tests

Fix doc comments

Avoid doing a `instanceOf` on a abstract type (which trigger a full type
check) and instead use a virtual function that create the basic type
which should be cheaper

Memoization depends on hash and equals. Otherwise we create duplicate memos for equal types

WIP: Make Mapping BTypes use semtypes

Fix initialization bug

WIP: add hack to make never? on the positive side work

Widened type for basic types is the basic type where as for other types it is contextually expected type (not the actual shape of the value). This is to avoid the overhead of having to create singleton types for each basic type value

Fix runtime type clash

Due to reasons that are not entierly clear to me when we run all the
unit tests some how types from `record_project_closed_rec_equiv` to
clashed with types defined in other unit tests. As a temperary
workaround I introduced a new package name for this.

Impelement xml semtype