fix: possible overflow in difficulty calculation (fixes #3923)

[jorgeantonio21]

Description

• First attempt to tackle issue Downcasting in difficulty calculation can lead to overflow  #3923

Motivation and Context

• The main goal is to avoid possible overflow of difficulty, which even though extremely unlikely, it is a possible scenario.

How Has This Been Tested?

• So far, changes have not been tested. I am more than happy to get recommendations of tests. I haven't add them at the moment because it is not clear how to build an hash whose big endian is a scalar less than 2^192. I am therefore, very interested to understand this concept and add tests accordingly.

[CjS77]

You can also write simple unit tests that are something like

#[test]
fn high_target() {
 let target: &[u8] = &[0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 ];
let expected = 1; # replace by the actual difficulty
assert!(big_endian_difficulty(target), expected);
} 

#[test]
fn max_difficulty() {
 let target = u64::MAX;
let expected = 1; # replace by the actual difficulty
assert!(big_endian_difficulty(target.to_be_bytes()), expected);
} 

#[test]
fn stop_overflow() {
 let target = 64;
let expected = u64::MAX; # replace by the actual difficulty
assert!(big_endian_difficulty(target.to_be_bytes()), expected);
} 

[jorgeantonio21]

#[test]
fn high_target() {
 let target: &[u8] = &[0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 ];
let expected = 1; # replace by the actual difficulty
assert!(big_endian_difficulty(target), expected);
} 

#[test]
fn max_difficulty() {
 let target = u64::MAX;
let expected = 1; # replace by the actual difficulty
assert!(big_endian_difficulty(target.to_be_bytes()), expected);
} 

#[test]
fn stop_overflow() {
 let target = 64;
let expected = u64::MAX; # replace by the actual difficulty
assert!(big_endian_difficulty(target.to_be_bytes()), expected);
} 

I added these tests, with minor mods to get the correct values. Thank you for the suggestion

[CjS77]

Looking good. If you could finlise the test cases to improve the coverage slightly, and not keep feature flags consistent, we should be good.

(And a note that the little_endian version has the same bug, but this is for another PR)

[CjS77]

I think you also need to run cargo fmt -p tari_core to pass the clippy tests here.

[CjS77]

LGTM - assuming tests pass