Benchmark PDU loop

[jamwaffles]

This is the most pure/basic possible benchmark of the PDU loop. It:

1. Prepares a FPWR frame with 8 bytes of payload data (32 bytes total) and marks it as sendable
2. Writes an Ethernet II packet containing the EtherCAT frame into a buffer
3. Passes the buffer to receive_frame which parses the data back out of the buffer

This test doesn't include any networking (not even loopback) because I'm on Windows so it's more a theoretical benchmark to check for regressions in the code than a real-world representation. On my Windows 11 (🤮) Ryzen 9 5950X system I get:

pdu-loop/elements       time:   [256.58 ns 257.65 ns 259.05 ns]
                        thrpt:  [3.8603 Melem/s 3.8812 Melem/s 3.8974 Melem/s]

pdu-loop/payload bytes  time:   [259.66 ns 261.35 ns 263.59 ns]
                        thrpt:  [28.944 MiB/s 29.193 MiB/s 29.382 MiB/s]

3.8m full frames/sec or 29MiB/sec EtherCAT payload of 8 bytes. I'm sure we can do better but this is a decent starting baseline.

I'll add a Linux benchmark when my test rig arrives which can involve sending traffic over the loopback interface for a more representative sample of the performance.