BEP-6: Initialising connections
Abstract: Currently, one must initialise a connection in two steps, first
construct the Connection object, then define the actual connections. In most
typical cases, the connections follow simple schemes that can be described in
one line. It would be more intuitive to do it in a single instruction.

STATUS: DONE (docs to write)

Examples:
C = Connection(G, H, 'V', random_connectivity(p=0.1, weight=1*mV))
or
C = Connection(G, H, 'V', weight=1*mV, p=0.1)

I would favor the latter syntax, which I find more straightforward.
The type of weight keyword tells what initialisation function should be
called:
* single value: connect_full (dense connection with identical weights)
* function: connect_full (functional meaning)
* matrix: connect (can be dense or sparse)

If the keyword "p" is specified, then the connect_random method is called.
Maybe it is not explicit enough though? proba? probability? sparseness?
I would go for sparseness, which is the most explicit. In that case it could be
added as a keyword to connect_random (alias for p).

This does not cover all the cases, but it is still possible to use the
connect methods in specific cases, while keeping an intuitive and efficient
syntax.
