Contact dedupe

Group contact-book entries that belong to the same person.

Each contact is a string "name|email|phone"; any field may be empty. Two contacts are duplicates if they have the same normalized email or the same normalized phone number. Duplication is transitive: if A duplicates B and B duplicates C, then A, B and C are all in the same group, even if A and C share nothing directly. Names are never used for matching (two different people can have the same name).

Normalization:

• Email. Trim spaces and convert to lowercase. Split at @; in the part before the @, delete the first + and everything after it (ann+news@x.com becomes ann@x.com). If there is no @, or the part before the @ is now empty, or the part after the @ is empty, the contact has no usable email.
• Phone. Keep only the digits. If fewer than 7 digits remain, the contact has no usable phone. Otherwise only the last 10 digits are used (all of them if there are fewer than 10), so +1 (415) 555-0134 and 415.555.0134 are the same number.

A contact without a usable email or phone simply cannot match through that field.

Input: contacts, a list of strings in the format above.

Output: a list of groups, one for every contact including those that match nobody. Each group is a list of 0-based indices into contacts in ascending order. Order the groups by their smallest index. If contacts is empty, return [].