Allow astropy.table.hstack to be used with TimeSeries objects

Currently `table.hstack()` doesn't work for `TimeSeries` instances as the expectation is that there is a `'time'` column. As there is one in both/all inputs those columns get renamed and thus won't be recognized for the output TimeSeries.

So the main questions is to how to handle time in hstacks. Shall we 1) keep all time columns, with all but the first renamed? 2) or shall we drop them, or 3) something different?

I have ideas how to proceed with either 1) or 2), but rather get some input from stakeholders first than jumping into implementing something that is undesirable.

ping @astrofrog @taldcroft

Start by reproducing the failure through astropy.table.hstack with two TimeSeries inputs and inspect how their 'time' columns are renamed. Review the surrounding table and TimeSeries behavior before choosing whether to retain, rename, or drop duplicate time columns; done means the project has an agreed behavior and hstack produces a valid, tested TimeSeries result.

Written by the indexing model from the issue text.