Improve pylama linter output handling

[favadi]

The current pattern for matching pylama output is:

'\v^.{-}:([0-9]+):([0-9]+): +%(([A-Z][0-9]+):? +)?(.*)$'

By default, pylama output format looks like:

samples/pylama.py:99:101 E501 line too long (189 > 100 characters) [pycodestyle]
samples/pylama.py:104:101 E501 line too long (189 > 100 characters) [pycodestyle]
samples/pylama.py:11:9  f-string is missing placeholders [pyflakes]
samples/pylama.py:12:101 E501 line too long (150 > 100 characters) [pycodestyle]

Notice the there is no : in the first part of the output. It looks
like the linter expects pylama to a non-default --format option, or
pylama changed the default format at some point. pylama support
--format json that will produce following output:

{
  "source": "pycodestyle",
  "col": 101,
  "lnum": 99,
  "etype": "E",
  "message": "line too long (189 > 100 characters)",
  "filename": "samples/pylama.py",
  "number": "E501"
},
{
  "source": "pycodestyle",
  "col": 101,
  "lnum": 104,
  "etype": "E",
  "message": "line too long (189 > 100 characters)",
  "filename": "samples/pylama.py",
  "number": "E501"
},
{
  "source": "pyflakes",
  "col": 9,
  "lnum": 11,
  "etype": "E",
  "message": "f-string is missing placeholders",
  "filename": "samples/pylama.py",
  "number": ""
},
{
  "source": "pycodestyle",
  "col": 101,
  "lnum": 12,
  "etype": "E",
  "message": "line too long (150 > 100 characters)",
  "filename": "samples/pylama.py",
  "number": "E501"
}

This commit introduces following changes:

• forces pylama command to always use --format json. Not sure if this
  will cause any issues.

• use json_decode instead of regex to parse the errors

[w0rp]

I just double-checked the changelog and the JSON format option was added late last year: https://github.com/klen/pylama/blob/develop/Changelog

To support this we'll have to check which version of pylama is being used and only use the JSON format if the version is new enough. Have a look at how we check the version for flake8 as an example.

[favadi]

@w0rp I added version check to support both older and newer version of pylama, please take another look.

Btw, I referred to flake8 check and this line looks suspicious. I think it will execute poetry --version and pipenv --version instead of poetry run flake8 --version and pipenv run flake8 --version respectively.