Makes the Python particle API read more like numpy and match Processing.

• Positional operands. p.apply(COMBINE, "velocity", "force", op="add"), like np.add(a, b, out=...). out= and options stay keywords. The old a=/b= keywords now error.
• Attributes are created on first write. No more declaring custom attributes up front. GENERATE has nothing to infer a size from, so it needs components=.
• create_particles(1000) / create_particles(shape), matching Java's createParticles(int) / createParticles(PShape). attributes= is still there if you want to pre-declare.

  p = create_particles(1000)
  p.apply(REDUCE, "velocity", out="speed", op=LENGTH)
  p.apply(COMBINE, "velocity", "speed", op=MUL)

All the examples are updated to the new forms.