Fix some edge cases in string and show for Arb and Acb

[Joel-Dahne]

For compact output it uses the condense option for the string representation. This gives a shorted output in general, but not in all cases. For some exact input, such as integers, you get a much more compact output by not condensing the output and instead just removing the trailing zeros. It is not always clear which option is better however, so now it just computes both and takes the shortest.

With this change the matrix Arb[1 1] is now displayed as

1×2 Matrix{Arb}:
 1.0  1.0

instead of as

1×2 Matrix{Arb}:
 1.00{…72 digits…}  1.00{…72 digits…}