Type inference: Simplify internal representation of type paths

[hvitved]

Adding a trailing . to the internal representation of type paths greatly simplifies the logic for constructing and deconstructing type paths.

DCA reports a small speedup.

[Copilot]

Pull Request Overview

This PR switches the internal representation of type paths to always include a trailing dot, which streamlines segment counting and concatenation logic and yields a measurable speedup.

• Replaced the lengthAtLeast2 helper and simplified length() to count trailing-dot separators via a regex.
• Refactored append and stripPrefix to pure string concatenation under the new trailing-dot convention.
• Updated singleton to emit its encoded type parameter with a trailing dot.

Comments suppressed due to low confidence (2)

shared/typeinference/codeql/typeinference/internal/TypeInference.qll:194

• [nitpick] The doc for append could note that both this and suffix include trailing dots, so no additional separator is added. Clarifying this will help maintainers understand the simplified concatenation.

/** Gets the path obtained by appending `suffix` onto this path. */

shared/typeinference/codeql/typeinference/internal/TypeInference.qll:216

• The revised stripPrefix logic should be covered by unit tests, especially edge cases like empty prefix or when prefix equals the full path, to guard against regressions.

TypePath stripPrefix(TypePath prefix) { this = prefix + result }

[Copilot]

The comment references count(this.indexOf(".")), but the implementation now uses regexpReplaceAll(...).length() to count separators. Consider updating the comment to clearly describe that it counts dots in the string.

Suggested change

	// `result = count(this.indexOf("."))`
	// but performs better because it doesn't use an aggregate
	// Counts the number of dots in the string by removing numeric characters
	// using a regular expression and calculating the length of the resulting string.

[paldepind]

Nice 😎