[FEATURE]: Date and datetime functions for DQL

Use case

It would be extremely useful to cast datetime types to other formats, such as year, month, or day. This way, we could generate meaningful aggregations on datetime types. Example scenario: a simple forum where users can make posts and comments. Each post and comment has a "created date" which is a datetime timestamp. We could then count all posts made in a month (or on a particular day)

Links to Discuss, RFC or previous Issues and PRs

https://discuss.dgraph.io/t/how-to-groupby-date/3040
https://discuss.dgraph.io/t/datetime-functions-are-needed/9694
https://discuss.dgraph.io/t/what-is-dgraph-lacking/16010/79?u=mrwunderbar666

Links to examples and research

For example see postgresql where you can cast a timestamp to year:

created_date::year
OR
date_part('year', created_date)

Current state

According to my knowledge the current workaround is to create separate fields for year, month, day (in the client application for DGraph) and then run aggregations on these self generated fields. Another solution is the math() function with since(): https://dgraph.io/docs/query-language/math-on-value-variables/

Solution proposal

Example query to aggregate on year:

q(func: has(created_date)) @groupby(YEAR(created_date)) {
    count(uid)
}

Other variant:

q(func: has(created_date)) @groupby(datetime(created_date, year)) {
    count(uid)
}

Additional Information

No response

Start with the linked discussions and the math-on-value-variables documentation, then compare the proposed YEAR(created_date) and datetime(created_date, year) queries with PostgreSQL's date extraction examples. Define the supported date parts, query syntax, and aggregation behavior before locating the relevant DQL implementation and tests. Done means the agreed functions work for the stated aggregation cases and are covered by tests.