Previously the CouplingMap class only supported graphs which were fully
connected. This prevented us from modeling potential hardware which
didn't have a path between all qubits. This isn't an inherent limitation
of the underlying graph data structure but was a limitation put on the
CouplingMap class because several pieces of the transpiler assume a path
always exists between 2 qubits (mainly in layout and routing). This
commit removes this limitation and also adds a method to get a subgraph
CouplingMap for all the components of the CouplingMap. This enables us
to model these devices with a CouplingMap, which is the first step
towards supporting these devices in the transpiler.

One limitation with this PR is most fo the layout and routing algorithms
do not support disjoint connectivity. The primary exception being
TrivialLayout (although the output might be invalid) VF2Layout and
VF2PostLayout which inherently support this already. This commit lays
the groundwork to fix this limitation in a follow-up PR but for the time
being it just raises an error in those passes if a disconnected
CouplingMap is being used. The intent here is to follow up to this commit
soon for adding support for SabreLayout, SabreSwap, DenseLayout,
and StochasticSwap to leverage the method get_component_subgraphs added
here to make them usable on such coupling maps.

Noise adaptive layout doesn't use a CouplingMap so we can't check for a
disconnected coupling map in it.

This commit renames the get_component_subgraphs() method to components()
which is much more consise name. At the same time this adds caching to
the return just in case building the component subgraphs is expensive to
compute we only need to ever do it once.

In a previous commit the distance() method was updated to handle
disjoint graphs correctly. Prior to this PR it was expected to raise
when a path didn't exist between 2 qubits by nature of the distance
matrix construction failing if there was a disconnected coupling map.
Since that isn't the case after this PR the error condition was
changed to check explicitly that there is no path available and then
error. However, there was an issue in this case and self loops would
incorrectly error as well when instead they should return 0. This commit
updates the error check to ignore self loops so they return correctly.

Co-authored-by: Kevin Krsulich <kevin@krsulich.net>

THis commit renames the CouplingMap.components() method to
connected_components(). It also adds an example to the docstring to
better explain what a connected component is.

The current reduce() error behavior of raising on trying to reduce to a
disconnected coupling map is being depended on in other locations. To
avoid a potentially breaking change this commit reverts the removal of
that limitation in the method. We can look at doing that in the future
independently of this PR because removing this specific restriction on
the reduce() method is not 100% tied to generally allowing disconnected
coupling map objects.