API: fix milliseconds accessor in Timestamp/DatetimeIndex

[jreback]

This results in an API change, so holding off till later

https://github.com/jreback/pandas/tree/milliseconds

[jorisvandenbossche]

@jreback This comes back to the datetime.datetime API (https://docs.python.org/2/library/datetime.html#datetime-objects) which stores only the microseconds (for the sub-second level).

So it is the choice between following that API as is done now (as we say Timestamp is a subclass and interchanable with it), or to deviate from it to provide a maybe more logical and full attribute support (millisecond, microsecond, nanosecond).

It is a bit the same discussion as we had with Timedelta

[jreback]

yep, its somewhat inconsistent now as nanoseconds returns not the total but up to 999 nanoseconds, while microseconds returns up to 999999, and milliseconds is undefined.

so let's think about it and maybe change in 0.16

[shoyer]

@jreback This is settled by decision on #9185, yes? I think we can close this as "won't fix".

[jreback]

yep