added methods of istriu and istril for Transpose and Adjoint arguments

[KlausC]

Performance optimization can be achieved. The following demonstrates the difference:

julia> @btime istriu(qrf.R)
  2.192 μs (0 allocations: 0 bytes)
true

julia> @btime istril(transpose(qrf.R))
  3.537 ms (1 allocation: 16 bytes)
true

julia> typeof(qrf.R)  # this was a 960x960 sparse matrix with 20% fill ratio.
SparseMatrixCSC{Float64,Int64}

# the follwing uses the new method:
julia> @btime istril(qrf.R')
  2.227 μs (1 allocation: 16 bytes)
true

There is a performance improvement of factor 1500 (3.54 ms vs. 2.23 µs).

[JeffBezanson]

LGTM. @andreasnoack ?

[andreasnoack]

The AbstractArray fallbacks strike again. Thanks for fixing this.