Fix minimum difficulty bugs

[teor2345]

Motivation

There are a bunch of bugs in Zebra's difficulty implementation, and the Zcash difficulty specifications also have some bugs.

Solution

New changes:

• Revise the testnet minimum difficulty code based on zcashd's implementation Testnet min-difficulty rules are incorrectly specified zcash/zips#416
  • ZIPs 205 and 208 are wrong
  • the min-difficulty rules change difficulty adjustment, not the difficulty filter
  • all blocks pass the default difficulty filter
• Turn expect("minimum difficulty activation is checked earlier in the test") into an Err if the rule is not active (and so the function returns None)

Tests for spec errors:

• Test that testnet min-difficulty rules follow zcashd Testnet min-difficulty rules are incorrectly specified zcash/zips#416
• Test implicit conversion of the minimum difficulty threshold to the compact representation Implicit conversion of the minimum difficulty threshold to the compact representation  #1277
  • covered by the genesis difficulty equality test

Reviewed changes:

• Refactor block target spacing into NetworkUpgrade methods
• Make ExpandedDifficulty Debug byte order match Hash
  • Bug fix on PR Reverse displayed endianness of transaction and block hashes #1171
  • Without this change, test failures are really confusing
• Round-trip the PoWLimit through the compact representation
  • the Zcash specification is ambiguous, see Implicit conversion of the minimum difficulty threshold to the compact representation  #1277
• Add minimum difficulty testnet block vectors
• Add tests for testnet minimum difficulty blocks
  • the Zcash specification is ambiguous, see Off-by-one error in testnet minimum difficulty block gap specification #1276
  • find a pre-Blossom and post-Blossom minimum difficulty block, and add them to zebra_test::vectors::block, along with their previous blocks
  • make sure the BlockVerifier difficulty tests pass on these new blocks

The code in this pull request has:

• Documentation Comments
• Unit Tests and Property Tests

Review

@yaahc has reviewed the Difficulty Contextual Validation RFC, which describes these checks.

Related Issues

Difficulty Contextual Validation RFC PR #1246
Difficulty tracking issue #802

Fixes bugs in:

• Previous difficulty check PR Block difficulty checks - context-free #809
• Reverse the displayed endianness of hashes Reverse displayed endianness of transaction and block hashes #1171
• Off-by-one error in testnet minimum difficulty block gap specification Off-by-one error in testnet minimum difficulty block gap specification #1276 (Zcash specification)
• Implicit conversion of the minimum difficulty threshold to the compact representation Implicit conversion of the minimum difficulty threshold to the compact representation  #1277 (Zcash specification)
• Testnet min-difficulty rules are incorrectly specified Testnet min-difficulty rules are incorrectly specified zcash/zips#416

Follow Up Work

• Move the difficulty filter check into the state service - obsoleted by Testnet min-difficulty rules are incorrectly specified zcash/zips#416
• allow minimum difficulty testnet blocks
• Change the difficulty threshold check into a method on Block or ExpandedDifficulty
  • it's easy to get <= confused with <

This work is listed in the difficulty tracking issue #802

[teor2345]

Unblocked by Daira, who said that blocks 299188, 299189, and 299190 are minimum-difficulty blocks.
That's enough info to move forward with this PR, and get it out of a draft state.

I also un-assigned yaahc as the reviewer, because this PR is not ready for review yet. I'm not sure if it actually fixes the bug.

[teor2345]

@yaahc after you approve this PR, I'd like to squash and rebase it myself, so we can keep some of the changes to different modules in different commits.