Apply multiple validation rules and return all failures

You will write a function validate_record(record) that accepts a dictionary representing a user-like record and returns a list of validation error messages. The validator must apply multiple checks for each field and report every failure found, not just the first. Order of messages must be deterministic so tests are reliable. Fields to validate: - username (required): string, 3-20 characters, only alphanumeric characters and underscores. - age (required): integer, between 0 and 120 inclusive. - email (required): string containing exactly one '@'; the domain part (after '@') must contain at least one '.'. - password (required): string, at least 8 characters, contains at least one digit, one uppercase letter, and one lowercase letter. - tags (optional): if present, must be a list of non-empty strings and must not contain duplicates. Additionally, any keys in the input not in the allowed set {username, age, email, password, tags} should be reported as unexpected fields. The function should return a list of error message strings in a fixed order (see examples). If there are no validation failures, return an empty list. Be defensive: ensure type checks are done before applying type-specific validations to avoid exceptions.

Implement a robust validator that applies multiple checks to a dictionary record and returns a list of all validation failures (no short-circuiting).

Do not raise exceptions for invalid input; collect and return validation messages. Keep message strings exactly as specified so tests can compare deterministically.