Add exists filter syntax

Is your feature request related to a problem? Please describe.

When querying for the existence of a one-to-many related model with certain values it one of its fields, the .distinct() method is required to prevent duplicates. This is not the most efficient query.

Contrived example: I have to do this to find all phone numbers of users that work at the north pole without duplicates:

PhoneNumber.filter(user__company__location='North Pole').distinct()

According to ChatGPT,

SELECT m.*
FROM model m
WHERE EXISTS (
    SELECT 1
    FROM user u
    JOIN company c ON u.company_id = c.id
    WHERE u.id = m.user_id
      AND c.location = 'North Pole'
);

is preferable over

SELECT DISTINCT m.*
FROM model m
LEFT OUTER JOIN user u ON m.user_id = u.id
LEFT OUTER JOIN company c ON u.company_id = c.id
WHERE c.location = 'North Pole';

Why this is better (copied from ChatGPT):

• EXISTS checks for existence rather than joining, so no duplicates.
• Often more efficient, since the DB can stop searching after the first match.
• Cleaner semantics: “Does this related row exist?” rather than “Give me all the related rows, then deduplicate.”

Describe the solution you'd like

PhoneNumber.filter(user__company__exists__location='North Pole')

Describe alternatives you've considered

Could be variations on the placement of __exists in the query. I realize the exact semantics could be a bit unclear.

Additional context

Related #385

Start with related issue #385 and compare the proposed filter syntax with the EXISTS and JOIN SQL examples in this issue. Clarify where __exists belongs and what its semantics should be; done means an agreed syntax that generates existence-based filtering without requiring distinct().