Proposed fix for Bug #15259 (invalid empty array returned by intarray "&" operator)

The attached patch fixes Bug #15259 [1] in the intarray module, making the
'&' array intersection operator return proper zero-dimensional empty arrays
instead of one-dimensional, zero-length "empty" arrays.

In [2] this problem was addressed by changing the behaviour of
construct_[md_]array(), but the intarray module does not use these
functions. The patch I propose contains the relevant fixes to the
intarray module, along with regression tests.

[1]:
https://www.postgresql.org/message-id/153053285112.13258.434620894305716755%40wrigleys.postgresql.org
[2]: https://www.postgresql.org/message-id/20570.1506198383@sss.pgh.pa.us

Best regards,

        Alexey Kryuchkov