Add support for finite data traversals

[oubiwann]

This will include something like setting an option / options such as:

• Traversal type #m(type random-neighbor)
• Point criterium:#m(start #(x0 y0) end #(x y))
• Count criterium: #m(count 100)
• Time criterium: #m(duration "M5") (see Erlang iso8601 and format info on Wikipedia)

Tasks:

• The implementation will use application state (loise-state) (Support centralised, long-running state data #64)
• A named layer will be operated against (Manage noise stacks / layers / bands #63)
• Dynamically spawn a child worker
• Use starting point + state data + traversal function to calculate next point
• Before running traversal function, check to see if one or more criteria have been met
• Store the points in the worker state
• A set of points will be returned that represent the path through the data
  • x,y coords
  • maybe also the name of the matrix that was operated upon?
• Before killing the worker, return the points

Depends upon:

• Provide a data API #55
• Manage noise stacks / layers / bands #63
• Support centralised, long-running state data #64
• Implement traversal stop criterium: duration #73

Blocks:

• Implement a random walk scan through noise data #53
• Add "linear scan" function #54
• Add a "downhill" noise traversal function #70
• Add an "uphill" noise traversal function #71