Fix TypeParser + TypeLexer

[instabledesign]

Hi,
This gonna to fix 1 wrong doctrine/lexer usage.
The wrong usage was located in the TypeLexer

    protected function getCatchablePatterns()
    {
        return [
            '\'(?:[^\']|\'\')*\'',
            '([a-z0-9\\\\]+)', // <----- You must not use parenthesis here
        ];
    }

You must not use group capture (parenthesis) because this regex was use in preg_split function
For example with array<key=int, value=string>
if you use ([a-z0-9\\\\]+) you get 15 tokens in the lexer

Array
(
    [0] => Array
        (
            [value] => array
            [type] => 2
            [position] => 0
        )
    [1] => Array
        (
            [value] => array
            [type] => 2
            [position] => 0
        )
    [2] => Array
        (
            [value] => <
            [type] => 5
            [position] => 5
        )
    [3] => Array
        (
            [value] => key
            [type] => 2
            [position] => 6
        )
    [4] => Array
        (
            [value] => key
            [type] => 2
            [position] => 6
        )
    [5] => Array
        (
            [value] => =
            [type] => 7
            [position] => 9
        )
    [6] => Array
        (
            [value] => int
            [type] => 2
            [position] => 10
        )
    [7] => Array
        (
            [value] => int
            [type] => 2
            [position] => 10
        )
    [8] => Array
        (
            [value] => ,
            [type] => 6
            [position] => 13
        )
    [9] => Array
        (
            [value] => value
            [type] => 2
            [position] => 15
        )
    [10] => Array
        (
            [value] => value
            [type] => 2
            [position] => 15
        )
    [11] => Array
        (
            [value] => =
            [type] => 7
            [position] => 20
        )
    [12] => Array
        (
            [value] => string
            [type] => 2
            [position] => 21
        )
    [13] => Array
        (
            [value] => string
            [type] => 2
            [position] => 21
        )
    [14] => Array
        (
            [value] => >
            [type] => 4
            [position] => 27
        )
)

and if you use [a-z0-9\\\\]+ you get 10 tokens in the lexer

Array
(
    [0] => Array
        (
            [value] => array
            [type] => 2
            [position] => 0
        )
    [1] => Array
        (
            [value] => <
            [type] => 5
            [position] => 5
        )
    [2] => Array
        (
            [value] => key
            [type] => 2
            [position] => 6
        )
    [3] => Array
        (
            [value] => =
            [type] => 7
            [position] => 9
        )
    [4] => Array
        (
            [value] => int
            [type] => 2
            [position] => 10
        )
    [5] => Array
        (
            [value] => ,
            [type] => 6
            [position] => 13
        )
    [6] => Array
        (
            [value] => value
            [type] => 2
            [position] => 15
        )
    [7] => Array
        (
            [value] => =
            [type] => 7
            [position] => 20
        )
    [8] => Array
        (
            [value] => string
            [type] => 2
            [position] => 21
        )
    [9] => Array
        (
            [value] => >
            [type] => 4
            [position] => 27
        )
)

so i fix the TypeParser::walk() function in order to drop the foreach (currently was a workaround to skip the token doublon capture by the regex parenthesis)

I'm not happy to set a lot of ->walk() call everywhere in the TypeParser but they are needed.

This PR is related to doctrine/lexer#12 (comment)
I've also tested with https://github.com/doctrine/lexer/pull/12/files#diff-3945e835e8d3a0ad8409023030a9db04R262

[instabledesign]

Tests fail look unrelated.

[instabledesign]

@egeloen can you check plz