TX/RX FSMs need refactoring

Nobody has claimed this yet.

These modules have a number of stylistic problems and poor documentation. Currently they do seem to work as intended but are something of a blackbox. Refactoring is worthwhile but low priority.

`axis_gmii_tx.sv`:
- The FSM transition logic is described by a `always @*` block that is over 200 lines and includes logic for driving outputs
- The some `logic` types are not reset
- Many flops sharing one processes
- Very little documentation for comments

`axis_gmii_rx.sv`:
- Inconsistent formatting, bad style, no comments
- Many `logic` types without reset
- Many flops sharing one process
- FSM transition logic and output logic described in the same `always @*` block
- Excessive reliance on processes over explicit combinational logic
- Not clear what the distinction is between `error_bad_frame`, `error_bad_fcs`, and `m_axis_tuser`

Start with axis_gmii_tx.sv and axis_gmii_rx.sv, especially their FSM transition and output logic blocks. Compare the reset, process-sharing, formatting, and commenting concerns listed for each module, and clarify the distinction among error_bad_frame, error_bad_fcs, and m_axis_tuser. Done means both FSMs are easier to understand, consistently documented and styled, and their reset and logic organization issues are addressed.

Written by the indexing model from the issue text.