coeff (correctly) not implemented for spoly used by various functions

[wbhart]

spoly cannot provide a coeff(f, n) function due to the fact that Singular does not have random access structures for polynomials.

This means the following are implemented incorrectly in Singular (or no longer needed):

canonical_unit(a::spoly{T}) -- src/poly/poly.jl
isnegative(x::spoly) -- src/poly/poly.jl

[wbhart]

In both cases coeff(f, 0) can be replaced with first(coefficients(f)), which will give the leading coefficient, which is presumably what was meant. Alternatively, perhaps leading_coefficient is better once the relevant PR is merged.

[fieker]

There is a chance that in Hecke at some point we (I) added hasLength to the iterator

[wbhart]

Just to clarify, addition of has_length wouldn't make these definitions correct, even if Singular depended on Hecke. Fortunately with the new printing code, these may not be needed any more anyway.