Regex caret (^) not working properly for scope: raw when there are over 999 characters in a Markdown file.

[michael-nok]

Check for existing issues

• Completed

Environment

• Windows 10
• Direct download of Windows executable
• Vale 2.29.1 or later

Describe the bug / provide steps to reproduce it

The changes implemented by this commit causes the problem to exist: f769fcd

	// NOTE: If the `ctx` document is large (as could be the case with
	// `scope: raw`) this is *slow*. Thus, the cap at 1k.
	//
	// TODO: Actually fix this.

I have a rule that looks for incorrectly indented content. It uses the following token:

extends: existence
message: 'Content must be indented using 4x spaces each time. "%s"'
level: error
nonword: true
scope: raw
tokens:
  - '^[ ]{1,3}\`'

When the Markdown file contains 999 characters or more (i.e. ctx > 1000), the ^ part of the token stops using the start of the line properly and invents (hallucinates) new starting positions.

Attached are sample files that exactly show the spillover in the logic:
vale.zip

Consequently, text with four spaces before the first ` is flagged as incorrect, and the starting position for the ^ is column 1.

Using the vale.exe from release 2.29.0 does not have this problem.