recalculation of LinkNode position after reloading a LinkProxy

[steinmet]

fragment of former #24:

Some remarks of Martin

• the `moving and jumping' happens for edge corners and LinkNodes?

• The reason might be, that for reloading a LinkProxy? sometimes the algorithm detects bogus invalid edges.

• The result of the effect above is, that it forgets about all the geometry information and generates a default `net' with the node placed into the gravity center of the connected terminals

A solution would be:

• to be more precise with the invalid edges
• to keep as much as possible information when invalidating a net (recalculate only the single edges that are invalid)

[rempel]

1) right now the update (reload) of net geometry is asynchronous (via scheduled update) which is necessary for moving linkNodes. this causes a delay for the update of the displayed link. i am not aware of any other moving and jumping. more specific test cases might be helpful to asses what needs to be fixed here.

2+3) the main case for invalid edges is that adding one connection invalidates all other link edges (which is unnecssary). these edges becomes valid again after reloading. rejection of these edges (because start and end index is -1 which should only happen for deleted edges) causes the loss of the link geometry, which is still contained and accurate after invalidation and generate of a default net.