Clean or Transform Data in Python

Phenom Talent Marketplace receives candidate skill labels from multiple sources. Write a function that cleans these labels, applies known aliases, removes duplicates, and ranks the resulting canonical skills by frequency.

Requirements

1. Normalize each skill by converting it to lowercase, replacing every sequence of characters other than letters, digits, +, and # with one space, and trimming surrounding spaces.
2. Normalize alias keys using the same rule. If a normalized skill appears in aliases, replace it with the normalized alias value.
3. Ignore empty or whitespace-only skill entries.
4. Count how often each canonical skill occurs, including occurrences that become equal after normalization or alias replacement.
5. Return canonical skill names ordered by decreasing frequency. Break ties alphabetically.

Formal Specification

Implement normalize_skills(raw_skills, aliases), where raw_skills is a list of strings and aliases is a dictionary mapping string aliases to canonical string names. Return a list of canonical skill strings. Do not modify either input.