Add server <-> server transfers of pdarrays

[bmcdonald3]

Add the ability to send a pdarray from one server to
another server through ZMQ sockets. This can be useful when
multiple Arkouda servers would like to operate on smaller
subsets of one giant dataset that can't fit on each Arkouda
server individually. For example, one large "mother" Arkouda
server could load in the entire dataset, down-select the data
as needed for each "child" server, then send the portion of
that array to the child server that would like to operate on
that portion of the dataset.

Prior to this PR, this would need to be done by writing the
down-selected portion from the mother server, reading those
files into the child server, and then operating. This aims
to replace the file system go-between of this use case,
simplifying the number of steps needed to get a pdarray
from one server to another and also potentially achieving
better performance than would be possible using the file
system.

Usage:

Server-1 ("mother" server):

>>> a = ak.read('my_file') # file that's too big for other servers
>>> b = a[0:int((a.size/2))] # get a portion of the array to send to child
>>> b.send_array("child-name", 1234) # send smaller array to server running on "child-name"

Server-2 ("child" server):

a = ak.receive_array("mother-name", 1234) # receive array from "mother-name"
                                    # port must match the port from the `send` call

Types supported in initial implementation with send_array/receive_array interface:

• all native Arkouda datatypes (int, uint, bool, float)
• categorical
• strings
• segmented array

Additionally, DataFrames containing columns of the above supported types can be transferred with a slightly different interface of DataFrame.transfer(hostname, port)/ak.receive_dataframe(hostname, port)