refactor apply_inputs and support converting block for v2 compatibility

[antiochp]

Refactor apply_inputs() and validate_inputs() to reduce duplicated code and simplify block conversion for future compatibility and v2 peer support.

Extracted some relatively isolated changes from #3385.

• utxo_view is now "source of truth" for identifying outputs being spent by inputs
• we now convert a block to v2 compatibility when processing new blocks
  • uses utxo_view.validate_inputs() to identify outputs being spent
• apply_inputs() now uses utxo_view.validate_inputs() internally
  • removed duplicate lookup code in apply_inputs()

We need to hook block conversion into the block processing pipeline early on, but after orphan block identification.
This required some refactoring of chain.process_block_single().
Checking for orphan block is simplified with an explicit check for "previous block in db".

This is a relatively large PR but mainly refactoring with the addition of the "v2 block conversion" step (lookup features for each output being spent).

We have good test coverage of the block processing pipeline.
Testing this on mainnet also to ensure the node syncs and stays in sync.

[antiochp]

Merging this - tested thoroughly on floonet and mainnet (and mining on usertesting).