Improve writing of Time and TimeDelta in "ascii.mrt" tables #20662

What is the problem this feature will solve?

Time and TimeDelta objects can be written to MIT tables, but all their meta information (like units) is lost. The code should set units for time deltas (and time where consistent with MRT standards, e.g. AAS journals user the units of days ("d") for JD, MJD, ...) and sensible default for the column descriptions.

Describe the desired outcome

Here is an example of current astropy behaviour:

import sys

from astropy.table import Table, QTable
from astropy.time import Time, TimeDelta
from astropy import units as u

times = Time(['2024-01-01T00:00:00', '2024-01-02T12:00:01'], format='isot', scale='utc')
times.format = 'mjd'
timediffs = TimeDelta([0, 1.500001] * u.day)
tab = Table({'obs': ['a', 'b'], 'time': times, 'timediff': timediffs})

tab.write(sys.stdout, format='ascii.mrt')

gives

Title:
Authors:
Table:
================================================================================
Byte-by-byte Description of file: table.dat
--------------------------------------------------------------------------------
 Bytes Format Units  Label     Explanations
--------------------------------------------------------------------------------
    1  A1     ---    obs      Description of obs     
 3-19  A17    ---    time     Description of time    
21-28  A8     ---    timediff Description of timediff
--------------------------------------------------------------------------------
Notes:
--------------------------------------------------------------------------------
a 60310.0           0.0     
b 60311.50001157408 1.500001

timediff doesn't have a unit, and time doesn't have any system. From the output, I cannot identify if time is in seconds since launch of the observatory, in MJD, in some other system and the 'time` column has about 6 more digits than it should have. One might say that it's up to the user to specify this in the "explanations" column, but astropy should set the units for the timediff and a sensible detail for the explanation of the time. (While the table writer can't know if this data was observed at the VLT or with HST, it could at list default to "Time in MJD".)

While an "astropy.time.Time" object doesn't have a unit, the AAS MRT example on https://journals.aas.org/mrt-standards/ explicitly lists a date in BJD with a unit of "d". I don't think there is a specified "you need to use BJD" or you need to use "ISOT" format in the MRT definition - it's up to the authors to pick what's best.

Additional context

PR #12027 had a draft implementation for this, but never got finished after the GSOC ended and was eventually closed by the bot.
While I've tried to summarize the discussion in that PR here, there are more details on how this might be implemented, which time format to choose etc.