feat: use 1 buffer for both SPI TX/RX ops

[2bndy5]

Opening this PR just to see if the compile size is reduced any. I haven't had a chance to properly test this.

[github-actions]

Memory usage change @ 525d5c6

Board	Flash	%	RAM for global variables	%
arduino:avr:nano	0 - 0	0 - 0	0 - 0	0 - 0
arduino:samd:mkrzero	0 - 0	0 - 0	0 - 0	0 - 0

Click for full report per board

arduino:avr:nano

Sketch	Flash	%	RAM for global variables	%
examples/GettingStarted	0	0	0	0
examples/AcknowledgementPayloads	0	0	0	0
examples/ManualAcknowledgements	0	0	0	0
examples/StreamingData	0	0	0	0
examples/MulticeiverDemo	0	0	0	0
examples/InterruptConfigure	0	0	0	0
examples/scanner	0	0	0	0
examples/encodeRadioDetails	0	0	0	0

arduino:samd:mkrzero

Sketch	Flash	%	RAM for global variables	%
examples/GettingStarted	0	0	0	0
examples/AcknowledgementPayloads	0	0	0	0
examples/ManualAcknowledgements	0	0	0	0
examples/StreamingData	0	0	0	0
examples/MulticeiverDemo	0	0	0	0
examples/InterruptConfigure	0	0	0	0
examples/scanner	0	0	0	0
examples/encodeRadioDetails	0	0	0	0

[2bndy5]

oh, right! the SPI buffers would only affect Pico SDK, Linux, and XMEGA_D3 platforms.

Ok, this doesn't seem worth pursuing now. All those platforms have the extra 33 bytes of memory to spare.

[2bndy5]

I'm going to clean up some other unused branches. Namely

• the interrupt branch from back in 2016. Although the only commit references Initializing pthread mutexes #193 which was closed by add static initialization for pthread mutexes #273
• c-port branch has no significant commits on the remote. I imagine @Avamander had local commits that never got pushed (circa 2017). Either way, I doubt that a port to pure C is feasible.