Improve some errors

[mcabbott]

This tries to make the errors from this package just a hair less inscrutable.

Edit: closes #81

[oxinabox]

LGTM except the think with isnothing

[oxinabox]

Oh sorry I missed this before, but shouldn't we also error if rule is not defined for frule?

[mcabbott]

Probably, I only fixed things which bit me, so far...

[mcabbott]

I've added one for frule too now. (And I think I got the signature right!)

BTW I'm not entirely sure MethodError is the right thing here. It has the right information, but isn't quite the truth -- a method does exist, it's just the fallback one. Maybe an explicit description is clearer?

[nickrobinson251]

closes #81

[oxinabox]

BTW I'm not entirely sure MethodError is the right thing here. It has the right information, but isn't quite the truth -- a method does exist, it's just the fallback one. Maybe an explicit description is clearer?

Manually thrown MethodErrors do that though.
E.g. https://github.com/JuliaMath/SpecialFunctions.jl/blob/master/src/gamma.jl#L528
I do agree sometimes they are confusing.
But the nice highlighting of what arguments are almost correct is I think worth it.

We can always change it again later