Make scanner accept floating point literals in hex and binary

[richcarl]

Similar to hex float literals in C 99, but using Erlang's Base#Literal notation we can easily have binary as well. For example, 2#0.10101p8, or 16#fefe.fefep-16. The letter p (as in C99) stands for powers of 2. The exponent, like the base, is always a decimal number.

These can be useful for code generators and other cases where you want to preserve an exact bit representation of a float.

Also allows explicit base 10 for completeness, as in 10#123.345e9 (with the letter e for exponent as normal - p not allowed, since it would make it very easy to miss whether it is a power of 2 or 10).

Currently missing: documentation. Might also need an EEP.

Based on #9104, which improves pinpointing of errors in numerical literals.

[github-actions]

CT Test Results

    2 files     96 suites   1h 7m 37s ⏱️
2 173 tests 2 125 ✅ 48 💤 0 ❌
2 536 runs  2 486 ✅ 50 💤 0 ❌

Results for commit 28ad2e8.

♻️ This comment has been updated with latest results.

To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass.

See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally.

Artifacts

• Complete CT logs (Download Logs)
• HTML Documentation (Download HTML Docs)
• Windows Installer

// Erlang/OTP Github Action Bot

[richcarl]

Rebased on #9104 after squashing.

[bjorng]

Looks good to me. We do need an EEP for any changes to the language.

[richcarl]

I'm seeing now that in Ada, which is where Erlang got its <Base> # <Digits> notation from, they have a slightly different notation for the exponent separator in a based numeral. Instead of p or P as in C99, they use #E. For example, 16#f.ff#E2 rather than 16#f.ffp2. It might be better to keep following the Ada style rather than creating a mix.