Fix float-mode build break: add MUL5A_SS / MUL6A_SS fallbacks

[rt-rtos]

What

Adds float-mode fallbacks for MUL5A_SS and MUL6A_SS in src/amy_fixedpoint.h.

Why

Since 98ea6bc ("Change multiply resolution to help Xanadu"), oscillators.c:18
references MUL5A_SS unconditionally:

#define MULA_SS(a, b) MUL5A_SS(a, b)

But MUL5A_SS and MUL6A_SS are defined only inside the fixed-point block of
amy_fixedpoint.h. Building with AMY_USE_FIXEDPOINT undefined (the float path
used on FPU targets — desktop, wasm/Emscripten, ESP32-S3, MicroPython) fails:

oscillators.c:18:23: error: implicit declaration of function 'MUL5A_SS'

Fix

Add the same (a) * (b) fallback that every other MUL*_SS macro already has
in the float block:

#define MUL5A_SS(a, b) ((a) * (b))
#define MUL6A_SS(a, b) ((a) * (b))

Notes

• The change is entirely inside #ifndef AMY_USE_FIXEDPOINT, so the default
  fixed-point build is untouched.
• Verified: the float translation unit reproduces the implicit-declaration error
  without these lines and compiles cleanly with them.

[dpwe]

Thank you so much!