Add MM_HALOW_ENABLE_LWIP to control lwIP integration.

[dpgeorge]

It's disabled by default. That allows the tests to be a lot simpler because they don't rely on lwIP anymore.

[dpgeorge]

@kwagyeman a small change, see what you think.

[kwagyeman]

From Claude:

Looks good — nice simplification, and defaulting MM_HALOW_ENABLE_LWIP off is
the right call.

One thing to confirm: is lwIP-off meant to be a buildable full-driver mode, or
just for the tests? The tests pass because they only compile
mm_halow_osal.c / mm_halow_sched.c, but mm_halow_ctrl.c still calls the
lwIP callbacks unconditionally (mm_halow_cb_tcpip_* /
mm_halow_cb_process_ethernet, around lines 134-151 and 442-462), and
mm_halow_lwip.c now compiles those out when lwIP is off — so a full driver
built with MM_HALOW_ENABLE_LWIP=0 would fail to link. If it's only for the
tests, all good; if it should build standalone, those call sites want the same
guard.

[dpgeorge]

is lwIP-off meant to be a buildable full-driver mode

Yes, eventually. Following cyw43-driver, I'd like to make a test that has a mock SPI backend and tests at least calling mp_halow_init(...).

I've updated this PR so it can at least compile all src files with MM_HALOW_ENABLE_LWIP disabled.

A follow-up PR can add a mock test that builds the whole thing.

[kwagyeman]

Okay, it's good. It's fine to merge. @dpgeorge

[kwagyeman]

From claude:

Yes — approve it.

Why it's good to approve:

It does what it claims and CI is fully green (formatting, spelling, host×2, qemu×2).
My review point was addressed head-on. Damien consciously scoped this as incremental: all src files now compile with lwIP off, and he explicitly deferred the full standalone link (the mock-SPI test that "builds the whole thing") to a follow-up PR. That's a reasonable maintainer call, not an oversight.

[dpgeorge]

Thanks!