Use iszero and isone in MPFR

[musm]

No allocations and is faster

julia> x = BigFloat(0)

julia> @btime iszero($x)
  9.474 ns (0 allocations: 0 bytes)
true

julia> f(x) = x == 0
f (generic function with 1 method)

julia> @btime f($x)
  99.874 ns (2 allocations: 56 bytes)
true

[Sacha0]

Perhaps nix the extra spaces and/or restore alignment with the preceding lines?

[musm]

I have another PR that changes this, which will conflict anyways. So I prefer to see one merge first before making any additional changes.

[Sacha0]

lgtm modulo the minor comments! :)

[giordano]

I don't have anything against this patch (actually, I'm in favor of it), but on Julia 0.6 I see a fairly different benchmark result:

julia> x = BigFloat(0)
0.000000000000000000000000000000000000000000000000000000000000000000000000000000

julia> @btime iszero($x)
  9.517 ns (0 allocations: 0 bytes)
true

julia> f(x) = x == 0
f (generic function with 1 method)

julia> @btime f($x)
  9.763 ns (0 allocations: 0 bytes)
true

(is BenchmarkTools still broken on master?)

[musm]

@giordano My benchmarks used the commit right before BenchmarkTools broke on master

[giordano]

On 952dc93, on my machine:

julia> x = BigFloat(0)
0.000000000000000000000000000000000000000000000000000000000000000000000000000000

julia> @btime iszero($x)
  9.453 ns (0 allocations: 0 bytes)
true

julia> f(x) = x == 0
f (generic function with 1 method)

julia> @btime f($x)
  9.167 ns (0 allocations: 0 bytes)
true

[musm]

maybe we should run the BigFloat benchmarks here

[andreasnoack]

@nanosoldier runbenchmarks("BigFloat", vs=":master")

[nanosoldier]

Your benchmark job has completed - no performance regressions were detected. A full report can be found here. cc @ararslan