Persistent modifier validation improvement

[oskin1]

Now we have quite flat validation algorithm for persistent modifiers - it does many checks in one step, this leads to impossiblity of handling all validation-failure cases properly. For example, if we're trying to validate block section (S) for some header (H) and validation fails at the first step then there are few cases which further application handling depends on:

1. S does not really relates to H - so we should just invalidate S
2. S relates to H but was malformed - here we should invalidate S and also H
   Current validation implementation in Scorex does not allow to distinguish case 1 and 2.
   https://github.com/ScorexFoundation/Scorex/blob/8adeddbaba7a49e0614f7df7de31fcaf73a263de/src/main/scala/scorex/core/NodeViewHolder.scala#L268

[oskin1]

Possible solution could be to subscribe to SyntacticallyFailedModification(S), check if some S relates to H it refers to, and then invalidate H.

[kushti]

@oskin1 what changes to Scorex you are proposing?