Improve performance for large number of portals about maxfield

I also used maxfield to plan a multifield of 104 portals.
It's current process.
44% (44/100 iterations) : 39h 15m 05.47s
Xeon(R) CPU E5-26xx 2.3GHz,DDR3

[tvwenger]

@pwiecz Can you confirm that your method creates the same number of fields as the current method? This algorithm was designed by @jpeterbaker to be optimal.

[pwiecz]

It creates the same number of fields.
I just don't understand the current algorithm well enough, to tell if it creates the correct link order (e.g. maybe edge dependency depends on the current triangulation algorithm).
I'm now trying to understand the code better (but I currently don't have much spare time :/ ).

[mvinni]

When I was developing the improveEdgeOrderMore stuff, I found automated validation of the generated plans very useful. To that end, you may find the last commit on this branch useful: https://github.com/mvinni/maxfield-1/tree/verify

This is the commit message:

add an option to validate the linking plan
    
    Use --check (or -c) to verify that the plan produced
    by the program does not contain some obvious errors.
    The checks are:
     - the source portal of a link should not be inside a field
     - a link can generate at most one field on each side

[mvinni]

I have been testing with an input of 119 portals, and managed to get the running time down from
about 2 minutes to 22 seconds with the exact same settings. I don't want to send a pull request before testing with even bigger inputs, but the current patches can be seen at https://github.com/mvinni/maxfield-1/commits/speedup

Maybe these are useful for others as well.