x25519/x448: audit and fixup Twist flag test cases

[cpu]

This is an extension of #196 with a few differences. Please check my work :-)

1. The x25519_test.json test case (tcId 101) that @XoifaiI flagged was also present in the x25519_pem_test.json and x25519_jwk_test.json variations. I amended their commit to fix these instances in the same manner as was done in x25519_test.json (Another point in favour of having less formats...)
2. I believe both the x25519 and x448 test cases had some acceptable result test cases that were missing the "Twist" flag.
3. I believe the x25519 test cases had some acceptable result test cases that had the "Twist" flag, but that weren't twist points.

I've included the program I wrote to lint these files, and integrated it with CI. This is mainly so others can check my work. I could be convinced it's not worth including the tool + CI integration. Please chime in if you have opinions here.

[XoifaiI]

Yep looks good, nice touch adding support for all test formats in twist check.
Very minor thing I can see is doing if legendre.Cmp(big.NewInt(0)) == 0 || legendre.Cmp(big.NewInt(1)) == 0 { rather than explicitly checking for -1, but its the same logically so I'd leave it.

And yeah I forgot forking the repo and then comparing would mean you can't push the change, I just did that so I could use the forks json link until it was fixed, thanks!

[XoifaiI]

Can confirm all the tests (including x25519) now pass for this now

[cpu]

I could be convinced it's not worth including the tool + CI integration. Please chime in if you have opinions here.

Will leave it for now. Easy to rm in the future if new opinions emerge.