feat: add sub-millisecond timestamp into uuid_generate_v7

[thme-tjpr]

This patch add a sub-millisecond precision to uuid 7, as provided in the rfc 9562.

Using an OPTIONAL sub-millisecond timestamp fraction (12 bits at maximum) as per Section 6.2 (Method 3). https://datatracker.ietf.org/doc/rfc9562/

Some applications need this kind of precision, especially when we include multiple records in batch and need to ensure that the ids are ordered by time.

Ex:
SELECT id, uuid_generate_v7() uuid from generate_series(1, 1000) as id order by uuid;

When we use 250*nanoseconds precision like in this patch, all the uuids are generated in sequence.

The original code return unsorted data like:
id uuid
19 01936e5d-d97e-700c-8dbc-23b886d51ffb
270 01936e5d-d97e-701d-91ac-0596339033cf
374 01936e5d-d97e-7020-bea1-0decee141a79
521 01936e5d-d97e-702e-bac9-9d83a0c35418

The patched code return sorted data like this:
id uuid
1 01936e62-33a0-7ea1-abb5-cc3197098a83
2 01936e62-33a0-7eaa-a8e5-59f6c28355e9
3 01936e62-33a0-7eb3-9aa1-f09dcffa352f
4 01936e62-33a0-7ebc-b099-a31e1d652135

[SakaSun]

+1
This is a nice feature to be added.

[reshke]

@fboulnois can you please take a look of it? It would be nice to generate uuidv7 in postgres < 18 (via this extension) exactly like in core(>=18)