[Request] writeBlocking(), writeFast(), & txStandBy() functions could be slightly faster

[2bndy5]

Currently, writeBlocking(), writeFast(), & txStandBy() functions have a while loop that fetches the status byte twice when waiting for an ACK response. If we save the status byte from every SPI transaction as a private member (uint8_t), we could simply refresh the status byte only when needed. This would reduce the polling of the SPI bus when waiting for a ACK packet.

The following source code snippet is from writeBlocking()

    while ((get_status() & (_BV(TX_FULL)))) {
        if (get_status() & _BV(MAX_RT)) {
            return 0;
        }
         ... // timeout algorithm removed for brevity
    }

Saving the status byte from all SPI transactions can reduce the need to refresh the status byte inside this while statement. There is an identical while loop in writeFast(). Both txStandBy() functions can also be optimized as there are similar while loops that fetch the status byte inside the loop.

Solution

Simply save the status byte from every SPI transaction as a private member. This should be fairly straight forward, and since there are several basic SPI functions that either return the status byte (write_*()) or ignore it (read_*()), I'm not going to get into a lot of detail here.

Alternative

If adding another byte worth of private member data is overstepping, we could instead use a locally scoped variable to save the status byte. This is the lazy way, and it doesn't open up further development opportunities (see below) concerning information from the status byte.

Additional context

Saving the status byte from every SPI transaction would actually be my first step towards implementing payload sizes on a per pipe basis as getting the payload size for the next available payload would require knowing which pipe received the payload (specifically concerning getPayloadSize())

Additionally I think the end-user would be better served to know if startWrite() & startFastWrite() were able to write to the TX FIFO at all by returning a boolean using the status byte's TX_FULL flag in bit position 0. Personally, I find it weird that this library has a rxFifoFull() but no txFifoFull(). This could be another issue for another time as I keep looking into the partially blocking write() functions.