More efficient 8b/10b line coding

[jvnknvlgl]

Currently the 8b/10b line coding uses two large (1024 and 2048 entry) lookup tables. This is not efficient, and can be done in a better way. Some of the suggestions when thinking about a new approach:

• Implement the full 8b/10b coding scheme in hardware (for reference, see FreeCores)
• Calculate the tables during compile time with TemplateHaskell

For all of these solutions, it would be nice to keep in mind that 8b/10b is a concatenation of 5b/6b and 3b/4b, and if it is implemented as such it greatly reduces the size of the lookup tables.

When implemented correctly, this also solves #2755