A common desire is to update bounds on variables based on some
additional source of data. When doing so, one wants to derive the
intersection of the existing bounds and the bounds following from the
additional data. When the types of these bounds do not match, this
becomes a difficult task, resulting in users setting bounds looser than
might be possible.

The convience method added in this commit handles such type conversions
for the user. It handles upcasting floats to Timeseries, and also
handles various combinations of vector bounds (as 2D Timeseries or
1D NumPy arrays).

Note that this convenience does come at the cost of performance. A
simple min/max merge on bounds which are known to be scalar can be done
in about 350 ns, whereas the method added in this commit takes about 10
times as long with 3.5 us. Merging a combination Timeseries and NumPy
arrays can easily take 30-40 us per call.

Closes #1074