Improve testing infrastructure

[KvanTTT]

Porting a PR from ANTLR4 repository (https://github.com/antlr/antlr4/pull/4301)

It provides several small goals instead of single big one:

• Support of separated lexer and parser in descriptors:

[grammar]
parser grammar Parser;

options { tokenVocab=Lexer; } 
...

[grammar]
lexer grammar Lexer;
...

It allows to port TestParserExec hardcoded tests to descriptors and check all runtimes instead of only Java.

• Simplify runtime test descriptors. No need to use type section because grammar type can be detected automatically (Lexer, Parser):

- [type]
- Parser
-

• Simplify tool tests. No need to specify grammar file, parser and lexer names because they are detected automatically:

ExecutedState executedState = execParser(grammar, "a", "x", true);

instead of

ExecutedState executedState = execParser("T.g4", grammar, "TParser", "TLexer", "a", "x", true);

• Add new checks on presence of generated files
• More clear architecture of test framework
• More than -1K lines of code without affecting any functionality. It's worth.

Also, it changes only testing infastructure but not core or runtimes. There are a lot of massive changes to test data because of removing redundant info.

It can be considered as follow-up of Runtime tests refactoring and Runtime tests refactoring 2.