der: support for the REAL type

[ChristopherRabotin]

Covers sections 8.5 and 11.3 of ITU-T X.690 (02/2021)

Caveats:
1. Interpretation of section 11.3.1 to be further discussed
2. Currently, all encoding is binary (no ISO 6093 NR3)

Testing of arbitrarily large and small f64s shows that DER typically
requires 12 bytes to encode these 8-byte IEEE-754 values. This can be
explained because of the extra two bytes for the tag and length, and the
fact that the exponent is encoded on 1 to 2 bytes (only 11 bits in
IEEE-754), and the mantissa may require up to 8 bytes itself, hence 12
bytes total. Of note that, depending on Caveat 1 above, this information
may be wrong.

Closes #304

[tarcieri]

If you can find another implementation to do interop testing with, that'd be great

[ChristopherRabotin]

Thanks for the review. I'll add a demo encoding from the ASN1 playground

[ChristopherRabotin]

As a quick update, I'm working on adding a bunch of validation test cases. It's gonna take some time as it seems like some of my encoding does not match the ASN1 playground so I'm going through the specs again to better understand what I didn't implement correctly. I'll keep the branch up-to-date with master.

[ChristopherRabotin]

After a two months of working on other stuff, here's the updated PR with validation cases. Let me know if you have any questions.

I'm not sure why clippy is failing in the Github actions, it works locally with clippy 0.1.59 (9d1b210 2022-02-23).

Cheers

[ChristopherRabotin]

Not sure what this CI build failed as running cargo test --no-default-features works locally and both the val and val2 variables are in fact defined.

[tarcieri]

It's failing here:

https://github.com/RustCrypto/formats/runs/5669154521?check_suite_focus=true

error: there is no argument named `val`
   --> der/src/asn1/real.rs:827:30
    |
827 |                 "fail - want {val}\tgot {val2}"
    |                              ^^^^^

error: there is no argument named `val2`
   --> der/src/asn1/real.rs:827:[41](https://github.com/RustCrypto/formats/runs/5669154521?check_suite_focus=true#step:7:41)
    |
827 |                 "fail - want {val}\tgot {val2}"
    |

I think the issue is that this is using the new captured identifiers in format strings feature which was introduced in Rust 1.58, and the MSRV of the crate is 1.57, which is what the test is failing on.

[ChristopherRabotin]

Fixed, thanks!

[ChristopherRabotin]

Thanks for the suggestion, that fixed all of the tests.

[tarcieri]

Thank you!