Prune list robustness

[antiochp]

We read the "prune list" from external data (via a txhashset download).
So we need to be a little careful with what we expect this to contain.

Specifically we make assumptions about this being 1-indexed and therefore will never contain a 0 value.

This PR adds some robustness around this - we still make the same assumption but we are not a little more defensive in how we deal with it.

• Use saturating_sub(1) in preference to - 1 to avoid overflows.
• Filter out 0 from our 1-indexed prune list when initializing the prune list.
• Add some assert! to make logic errors more visible if/when we misuse the prune list

[antiochp]

We've tested this for a bit in various scenarios.
Going to merge this.