Add methods to help with potential problems from #5556 - #5586

@taldcroft - Thanks!

I'm confused. In the description above you write:

Revert to the previous "always inplace" behavior for now, with deprecation and then removal of this option in subsequent releases.

In docs/table/modify_table.rst you write:

Starting with astropy 1.3 the operation shown above does a complete replacement of the column object.

So is the new default now the same as before, or is it replace for 1.3?

Otherwise, I think the new extended description of the change is great. Thank you for writing it up!

There's one thing that always drives me mad: it's pretty hard to figure out how to get a traceback that shows where warnings come from. Giving that info how to get a filename / line number in the "Finding the source of problems" would be helpful IMO. (could also be a more general part of the docs and then link to it).

@cdeil - setting the new configuration option table.conf.replace_inplace = True will revert to the pre-1.3 behavior (in-place always), however the default is False and so will have the new (replace-by-default) behavior.

OK, thanks for clarifying it for me!

@cdeil - I updated the description to make it more clear.

FYI I will be on travel and mostly offline from Monday 7pm EST to Friday 9pm EST.

@taldcroft - I think this all looks good (I'm not completely sure it is necessary, but certainly good). One worry is whether by default giving warnings on replacing a slice is going to give a lot of false positives. But I guess we will find out. Anyway, no objection to merging.

@mhvk - the halotools regression (astropy/halotools#637) was a little tricky to diagnose even after we knew that it was due to #5556 (which was possible because @aphearin has astropy-dev connections). So I would be very worried about random person out there getting totally frustrated that their code is failing in an obscure way.

Hopefully not too many false positives, but people can always disable the warnings.

OK, that was definitely a tricky one. 👍 on the slice warnings. And, on a bit more thought about reversion, continued 👍 on the change to replacing columns.

Alright, I'm going to merge this and backport to 1.3 so it gets into RC1. But I think we should keep an open mind about possibly changing the replace_inplace default to True for 1.3, and then switch to False in 2.0, depending primarily on whether or not we get complaints during the RC.

backported to v1.3.x in 7b69bad