Add header_align argument

[rlabbe]

The idea here is to be able to align the header separately from the values. For example, when I print floats I like the decimal points to align, and I get that with align='r'. But then the header names are all the way to the right, and you end up with

                   x
    12346.657

whereas I'd like:
            x
    12346.657

[hugovk]

I'd be open to looking at a PR for this.

[hugovk]

Re-opening as #183 was reverted, a new PR is welcome, thanks!

[myheroyuki]

Hi, I'm willing to give this one another attempt. @hugovk, what do you think the correct behavior should be. I think that header_align should take precedence if both header_align and align are set. It seems strange to me to allow the user to use align for a individual header fields when header_align exists, but I can see from the issue mentioned in the pull request it's needed to ensure backwards capability.

I think the best solution is that align should work as always as long as no header_align is set. If the user ever does set header align, it will be used to set the alignment for header fields. This could be bit confusing though if header_align is set very early on, and the user later changes a specific header field's alignment at the end using align, as that align value would be ignored. I suppose that's fine as long as it's documented though.

[hugovk]

Hi, thank you! Yes, that sounds about right :)

• If only align is set, it's the "global" option that applies to both header and body.

• If only header_align is set, use header_align for the header, no change to the body.

• If both header_align and align are set, use header_align for the header, and align for the body.

Ping also @osfrickler and @SeanMooney: how does this sound?

[SeanMooney]

"""If only align is set, it's the "global" option that applies to both header and body."""

i think that would maintain the behaviour we are currently depending on in our test yes.

the rest also make logical sense to me.