Reject blocks that violate turnstile
This is an implementation of a consensus rule which marks blocks as invalid if they would lead to a turnstile violation in the Sprout or Shielded value pools. The motivations and deployment details can be found in the [accompanying ZIP draft](https://github.com/zcash/zips/pull/210).
**This PR only introduces the rule for testnet at the moment.**
We achieve the institution of this rule in three ways:
1. Nodes prior to #2795 did not record the "delta" in the Sprout value pool balance as part of the on-disk block index. This was a long time ago, though, and all nodes that are consensus-compatible with the network today have been recording this value for newer blocks. However, the value is absent from older block indexes unless the node has reindexed or synchronized from scratch in the time since. We shouldn't need to require nodes to reindex in order to enforce this consensus rule. We avoid this problem by falling back on a hardcoded Sprout shielded value pool balance in a very recent block.
2. If during `ConnectBlock` we observe that the resulting shielded value pool balance of Sprout or Sapling is negative, we reject the block.
3. During the miner's block assembly process the miner will skip over transactions if adding them to the assembled block might violate the turnstile, since the resulting block would be invalid. This means that theoretical transactions violating the turnstile would still be relayed in the network (and made available in users' memory pools) and so a turnstile violation would have some visibility outside of block relay.
## Smoke Testing
It's really tricky to test the behavior that automatically falls back to hardcoded shielded value pool balances in our architecture because it's very testnet-specific and node-version-specific. However, we can do some smoke tests to see that everything is working.
I modified the serialization of `CDiskBlockIndex` to serialize `boost::none` for `nSproutValue`
```
if ((s.GetType() & SER_DISK) && (nVersion >= SPROUT_VALUE_VERSION)) {
boost::optional<CAmount> nSproutValueFake = boost::none;
READWRITE(nSproutValueFake);
}
```
and then began a reindex of my node which I interruped around height 130k on testnet. I then restored the original serialization and resumed the reindex; I have thus _roughly_ simulated a older node "upgrading" to a newer node that records the deltas when processing new blocks. My node showed pool monitoring was disabled, as expected, for Sprout. I confirmed that some blocks following the reindex had nonzero Sprout `valueDelta` from `getblock`, as expected. I finished the reindex, restarted the node, and confirmed that the serialization worked for newer blocks but not older blocks by querying `getblock`, simply as a reassurance.
Finally, I introduced the code in this PR and reloaded the node. The desired behavior (that the chain began to be "monitored" again) worked, and the values were consistent with the hardcoded constant. I then made a payment to a Sprout z-addr from the transparent pool and the pool value increased as expected, as reported by `getblockchaininfo`. I reindexed the node again to exercise the remaining logic and check for turnstile violations throughout the history of testnet; there were none.
Correcting logo on README
As a follow up to https://github.com/zcash/zcash/pull/3509 which I closed due to the updated branding. The logo [still has the wrong aspect ratio]. This corrects it with the logo taken from https://z.cash/press/ while preserving the original image dimensions and aspect ratio. I added it at 2x for retina displays.
This sidesteps the problem where the atomics check tries to run a test
binary, which cannot be performed during cross compilation. We should
replace this with a better solution in future.
Part of #3710.
Usage on Debian / Ubuntu:
> $ sudo apt install g++-aarch64-linux-gnu
> $ HOST=aarch64-linux-gnu ./zcutil/build.sh
Currently fails to cross-compile due to later configuration issues in
the depends system that need to be worked around.