Add caching for symbols retrieved by TokenTypeAnalyzer and SymbolReferenceAnalyzer

[costin-zaharia-sonarsource]

Both TokenTypeAnalyzer and SymbolReferenceAnalyzer retrieve the same symbols and by caching them we will reduce the processor time used by these 2 rules in half. I did a POC and for the repro I've added I have the following results:

from:

    31.131   28      SonarAnalyzer.Rules.CSharp.TokenTypeAnalyzer
    31.065   28      SonarAnalyzer.Rules.CSharp.SymbolReferenceAnalyzer

to:

    29.596   41      SonarAnalyzer.Rules.CSharp.TokenTypeAnalyzer
     0.019   <1      SonarAnalyzer.Rules.CSharp.SymbolReferenceAnalyzer

[martin-strecker-sonarsource]

@costin-zaharia-sonarsource I think this is outdated and can be closed:

• TokenTypeAnalyzer queries way less symbol info than before (<10% of what it used to query).
• Registration of the analyzer changed to "RegisterSemanticModel" which gives the same model for both analyzers and the model does caching on its own.

[costin-zaharia-sonarsource]

Thanks, indeed this has been changed in the meantime and the issue is no longer up to date.