…ception in runtime

…iables within a block) immediately

* Before Prism it worked thanks to __FILE__ returning the real path
  instead of `<internal:core> core/struct.rb` which is more correct.

* This avoids:
  Invalid source section bounds. (java.lang.IllegalArgumentException)
    from com.oracle.truffle.api.source.Source.assertValid(Source.java:735)
    from com.oracle.truffle.api.source.Source.createSection(Source.java:669)
* We need to remove the eager check using toJavaStringOrThrow() because
  e.g. spec/ruby/language/regexp/encoding_spec.rb fails it.
  .toJavaStringUncached() will anyway not fail but use � in that case.

…ception message

…rators

The original error to parse literal: "ArgumentError: invalid value for Integer(): 0x0000_0000_0000_0021"

In case at least one literal has forced encoding - we should involve encoding negotiation.

Code that reproduces the issue:

```ruby

accents = "" "\u0300"
```

Class/module constant source location could be retrieved with Module.const_source_location

Prism doesn't emit such warning yet

… Array argument

…ent#declareVar where possible

Prism emits different error and warning messages

… keywords

* And load the library when creating the first RubyContext.

* Until ruby/prism#2082 is resolved.

* So there is a Ruby home available.

…bleNode

* Revealed by TCK tests failing like this:
  java.lang.IllegalStateException: WrapperNode implementation org.truffleruby.language.RubyNodeWrapper
  cannot be safely replaced in parent node class org.truffleruby.core.rescue.AssignRescueVariableNode.

…f tags

* Basically the same as the TCK instrument (VerifierInstrument) but can be
  used to run any code vs only TCK code. Instrumentation is applied on the
  first execution, so very little code is instrumented during the TCK.
* This revealed several incorrect wrappers.

… check for it

* Running $ jt test fast -- --instrument-all-nodes

* See ruby/prism#2082

…ays work

* Because not every Ruby Encoding has a corresponding Java Charset.

* So we don't rely on Charset which does not exist e.g. for ISO-8859-10.

* Otherwise coverage does not work because instrumentation events
  are not called at all as it goes
  wrapper.doExecuteVoid -> node.doExecuteVoid -> node.execute
  instead of:
  wrapper.doExecuteVoid -> wrapper.execute (events) -> node.execute

* It leads to much cleaner and more robust code, as our ParametersNode fields are then never null.

It seems it's caused by the issue in ruby/prism.

Original exception:
```
_1 (java.lang.AssertionError)
	from org.truffleruby.parser.YARPTranslator.visitLocalVariableReadNode(YARPTranslator.java:2461)
	from org.truffleruby.parser.YARPTranslator.visitLocalVariableReadNode(YARPTranslator.java:181)
```

The following code
```ruby
h[*k], = ["ok", "ng"]
```
leads to failed assert that takes the last argument (placeholder) in a wrong way.

Original error:
```
ArrayConcatNode@e424bba at no source section (java.lang.AssertionError)
	from org.truffleruby.language.dispatch.RubyCallNode.assign(RubyCallNode.java:147)
	from org.truffleruby.core.array.MultipleAssignmentNode.assign(MultipleAssignmentNode.java:66)
	from org.truffleruby.core.array.MultipleAssignmentNode.execute(MultipleAssignmentNode.java:56)
	from org.truffleruby.language.RubyNode.doExecuteVoid(RubyNode.java:64)
	from org.truffleruby.language.control.SequenceNode.execute(SequenceNode.java:34)
	from org.truffleruby.language.RubyMethodRootNode.execute(RubyMethodRootNode.java:65)
/b/b/e/main/test/mri/tests/ruby/test_assignment.rb:937:in `test_massign_aref_lhs_splat'
```

…ntax.rb

* See ruby/prism#2287

* The untagged spec is legitimately passing due to Prism returning the correct SyntaxError for it.

* But with several cleanups and fixes.
* Also allow Prism error messages in pattern matching specs.

…a node

* So it can still be caught in user code, e.g. for missing find pattern matching.

* Keeps the exclusion tags in consistent order.
* More reliable and easier to reproduce.