Add counters to stm32xx-spi-server-core ringbuf

[hawkw]

The stm32xx-spi-server-core ringbuf fills up pretty quickly when the system is busy, so it's hard to see the total number of SPI operations that may have occurred in the past. This branch adds counters to this ring buffer.

Note that this PR depends on an upstream idol change, oxidecomputer/idolatry#49, to allow us to derive the Count trait for the SpiOperation generated type. So, we should merge that first, and get this pointed back at oxidecomputer/idolatry@main.

For example (on my gimletlet):

$ humility ringbuf spi_server_core
humility: attached via ST-Link V3
humility: ring buffer drv_stm32h7_spi_server_core::__RINGBUF in net:
   TOTAL VARIANT
    3374 WaitISR
    3184 Tx
    3184 Rx
     590 Start(exchange)
     206 Start(write)
 NDX LINE      GEN    COUNT PAYLOAD
   6  467      117        2 Tx(0x0)
   7  531      117        2 WaitISR(0x40012)
   8  501      117        3 Rx(0x0)
   9  531      117        1 WaitISR(0x20012)
  10  531      117        1 WaitISR(0x10012)
  11  501      117        1 Rx(0x40)
  12  359      117        1 Start(exchange, (0x2, 0x4))
  13  467      117        1 Tx(0x2)
  14  467      117        1 Tx(0xcc)
  15  467      117        2 Tx(0x0)
  16  531      117        2 WaitISR(0x40012)
  17  501      117        2 Rx(0x0)
  18  501      117        1 Rx(0x56)
  19  531      117        1 WaitISR(0x20012)
  20  531      117        1 WaitISR(0x10012)
  21  501      117        1 Rx(0x9)
  22  359      117        1 Start(exchange, (0x2, 0x4))
  23  467      117        1 Tx(0x5)
  24  467      117        1 Tx(0xb0)
  25  467      117        2 Tx(0x0)
  26  531      117        2 WaitISR(0x40012)
  27  501      117        2 Rx(0x0)
  28  501      117        1 Rx(0x8)
  29  531      117        2 WaitISR(0x10012)
  30  501      117        1 Rx(0x78)
  31  359      117        1 Start(exchange, (0x2, 0x4))
  32  467      117        1 Tx(0x5)
  33  467      117        1 Tx(0x8c)
  34  467      117        2 Tx(0x0)
  35  531      117        2 WaitISR(0x40012)
  36  501      117        2 Rx(0x0)
  37  501      117        1 Rx(0x20)
  38  531      117        2 WaitISR(0x10012)
  39  501      117        1 Rx(0x31)
  40  359      117        1 Start(write, (0x4, 0x0))
  41  467      117        1 Tx(0x83)
  42  467      117        1 Tx(0xc)
  43  467      117        1 Tx(0x20)
  44  467      117        1 Tx(0x1c)
  45  531      117        2 WaitISR(0x40012)
  46  501      117        2 Rx(0x0)
  47  531      117        1 WaitISR(0x22017)
  48  501      117        1 Rx(0x20)
  49  531      117        1 WaitISR(0x10012)
  50  531      117        1 WaitISR(0x301f)
  51  501      117        1 Rx(0x31)
  52  359      117        1 Start(exchange, (0x2, 0x4))
  53  467      117        1 Tx(0x2)
  54  467      117        1 Tx(0xf0)
  55  467      117        2 Tx(0x0)
  56  531      117        2 WaitISR(0x40012)
  57  501      117        3 Rx(0x0)
  58  531      117        2 WaitISR(0x10012)
  59  501      117        1 Rx(0x40)
  60  359      117        1 Start(exchange, (0x2, 0x4))
  61  467      117        1 Tx(0x2)
  62  467      117        1 Tx(0xcc)
  63  467      117        2 Tx(0x0)
   0  531      118        2 WaitISR(0x40012)
   1  501      118        2 Rx(0x0)
   2  531      118        1 WaitISR(0x20012)
   3  501      118        1 Rx(0x0)
   4  531      118        2 WaitISR(0x10012)
   5  501      118        1 Rx(0x0)
humility: ring buffer drv_stm32h7_spi_server_core::__RINGBUF in sprot:

[cbiffle]

I'd be kind of inclined to make this just counters, but that's because the ringbuf has come up in my profiles as a bottleneck for the SPI driver repeatedly. It's possible people are using its contents and converting it to just counters would be disruptive. Not sure.

[hawkw]

I'd be kind of inclined to make this just counters, but that's because the ringbuf has come up in my profiles as a bottleneck for the SPI driver repeatedly. It's possible people are using its contents and converting it to just counters would be disruptive. Not sure.

Hmm, yeah...adding counters to a ringbuf that already exists does seem like a less controversial change than changing a ringbuf to counters. But, we could do that as well, if no one feels like having the ringbuffer is important...