Add SPI loopback test app

[ppannuto-claude]

Summary

Adds examples/tests/spi_loopback, a small standalone app that exercises the
current libtock/peripherals/spi_controller.h API end-to-end: writes a known
pattern and reads it back in the same transfer, then checks the two match,
printing SPI PASS or SPI FAIL: <reason>.

This fills a gap found while adding SPI support to a new board in
tock/tock#5115: there was no
example that builds against the current libtock_spi_controller_* API — the
existing wip/spi/* examples predate it (different function names, no
returncode_t) and don't compile against it without a rewrite.

The test is only meaningful on a board where SPI is configured in hardware
loopback mode (read data is expected to equal write data), such as the new
QEMU qemu_arm_mps2_an385/qemu_arm_mps2_an386 boards. On a board wired to
a real, non-looped-back peripheral, a mismatch is expected and not a bug —
noted in the app's README.

Testing

Built for both cortex-m3 and cortex-m4 (TOCK_TARGETS=cortex-m3 make /
TOCK_TARGETS=cortex-m4 make), and run under QEMU against the boards in
tock/tock#5115 (qemu_arm_mps2_an385/an386) alongside c_hello and
blink, confirming SPI PASS. Re-verified SPI PASS again from the
current examples/tests/spi_loopback path after the review round below.

make format (uncrustify) passes clean with no changes.

AI disclosure

Written by Claude (Claude Code), as part of the same session that produced
tock/tock#5115, in response to:

❯ Clean up the SPI test app and open the libtock-c PR
❯ Address feedback on the libtock-c PR

The second prompt was in response to Pat's own review (as ppannuto,
requesting changes): moved the app to examples/tests/spi_loopback,
renamed wbuf/rbuf to write_buf/read_buf, dated the
set_chip_select comment, lifted the SPI rate to a named SPI_RATE_HZ
const with a comment on its provenance, printed both buffers on mismatch,
and dropped a redundant README note — see inline replies on the review for
detail on each.

Review status: reviewed once by Pat Pannuto (changes requested); the above
addresses that round. Not yet re-reviewed/approved.

[ppannuto]

Meta: This should move to examples/tests/spi_loopback/

[ppannuto-claude]

Moved to examples/tests/spi_loopback/ and addressed the inline review comments (renamed buffers, dated the set_chip_select comment, lifted the SPI rate to a named SPI_RATE_HZ const with a comment on its provenance, printed both buffers on mismatch, dropped the redundant README note). Amended the commit and force-pushed; re-verified SPI PASS under QEMU from the new path and re-ran make format clean.

[ppannuto-claude]

Fixed: print_buf now takes an explicit len parameter instead of using the global BUF_SIZE. Amended and force-pushed; rebuilt for both targets and re-confirmed SPI PASS under QEMU.

[ppannuto-claude]

Addressed the third review round: read_buf now pre-fills with i + BUF_SIZE (disjoint from write_buf's pattern) so a no-op can't false-PASS; set_chip_select's return is now checked against RETURNCODE_ENOSUPPORT explicitly; added the suggested progress print before yield_for. Amended and force-pushed, rebuilt for both targets, re-confirmed SPI PASS under QEMU.