Fix coverage simplifier issues 953 962

[mukoki]

Replace the previous pull request #961 : fix both issues #953 and #962 according to the discussion.

Fixes #953. Fixes #962.

[mukoki]

I tried to follow your suggestions about identification of nodes on shells sharing only one point. Finally, this new pull request fix several bugs :

• fix computation of CoverageEdge key in the case of ring including a single node (= first commit of my first PR)
• identifiying single node when two shells touch each other (propagate the information to CoverageEdge and Edge as a boolean) and pass it through MultiLineString via a BitSet included in the Geometry
• also fix PriorityQueue of corners which did not include all possible corners (I had to change the test testFilledHole() as the expected polygon was not simplified to the maximum before this change). Also clean the Queue from deprecated corners)
• add some tests and fix GeometryTestCase which was displaying the wrong message
• there is also an unfinished change in LinkedList (not sure it is necessary)

[mukoki]

On a significative dataset (117 Mb), the processing time goes from 41.5 s to 46.5 s after applying the PR.

[dr-jts]

You'll need to rebase, since I pushed a fix to GeometryTestCase.

[mukoki]

Sorry, with the different commits, sync and rebase, I'm not sure the PR is still usable for you. Let me know. I've also made a separate PR where I replaced the BitSet trick by a second MultiLineString to pass the unconstrained rings. Use the one you prefer or another way if you have a precise idea in mind.

[dr-jts]

This PR looks good, with one small change request as noted in a comment (pass the BItSet as an explicit parameter to TPVWSImplifier).