fix shape returned from keplerlib.propagate

[Tontyna]

fiixes #959

[brandon-rhodes]

@JoshPaterson — I'm inclined to accept this pull request, maybe after requesting that a test be written so that a future well-meaning coder doesn't break this edge case again. But I thought I should loop you in, since I think this is your code? Simply comment here over the next few days if you'd like to weigh in. Otherwise, if I don't hear from you by maybe the middle of next week, I'll assume you're busy these days. I hope you're doing well!

[JoshPaterson]

Hi @brandon-rhodes, I'll take a look at this. I'm wondering if instead of checking if squeezing will make the wrong shape and then using a conditional to fix it, can the squeezing be made to work correctly the first time? If either you or @Tontyna have ideas about that let me know!

[JoshPaterson]

Perhaps this fix can be accepted to get things working and then if it can be improved that can be a separate PR.

[Tontyna]

I tried to avoid reshape, but squeeze can't do nothing about it.
The whole calculation -- atleast_ld ! -- doesn't distinguish between a given scalar time and a [time], so we must remember (and reshape).

[Tontyna]

One could do without the expected_shape variable by not modifying the t1 parameter within the function and doing the shape-check / reshape only at the end of the function.