Come up with a better way to get the scalar

[eric-wieser]

As discussed in #371 (comment).

We find ourself using 1 + 0*x a lot of the time to obtain the scalar element.

Right, now, there is no great alternative to this spelling. The obvious choice of x.layout.scalar results in a multivector of int type, which is fine for most code, but is awkward for jitted code where really we want a scalar with the same coefficient type as x.

[hugohadfield]

Maybe x.unit_scalar or something similar might be a good spelling

[eric-wieser]

cf.one_like(x) or cf.scalar_like(x) is another option, sort of like np.zeros_like.