Carbon lexer doesn't handle multi-character integers correctly

[davidben]

I was looking at the Carbon lexer as a template for another one (it was very small and C-like), and noticed an issue.

#2370 (CC @AmrDeveloper) switched the integer regexes to +?, which is non-greedy. As a result, multi-character integers such as 1234 or 0x1234abcd get lexed funny. They're not covered by tests, but the lex ends up being:

'   '         Text.Whitespace
'1'           Literal.Number.Integer
'2'           Literal.Number.Integer
'3'           Literal.Number.Integer
'4'           Literal.Number.Integer
'\n'          Text.Whitespace

'   '         Text.Whitespace
'0x1'         Literal.Number.Hex
'2'           Literal.Number.Integer
'3'           Literal.Number.Integer
'4'           Literal.Number.Integer
'abcd'        Name.Other
'\n'          Text.Whitespace

[AmrDeveloper]

Thank you @davidben for reporting,
I will fix it and cover it on the test cases