Cache parsing of file stem for the same entity part

[blond]

I think we can slightly speed up scanning if will cache parsing of file stem (basename without extension) for the same entity part.

Example:

button/
  __text/
    button__text.examples/
    button__text.tmpl-specs/
    button__text.gemini-bundles/
    button__text.css
    button__text.js
    button__text.deps.js
    button__text.bemhtml.js
    button__text.bh.js
    button__text.spec.js
    button__text.ru.md
    button__text.en.md

The button__text entity part has 11 techs. Each tech can be written in several levels.

This means that we will be scanning 11 (or more) files with button__text stem.

Now for each file will be called the bemNaming.parse() method.