Faster boundedness check for invertible LinearMap

[schillic]

In #956 we implemented isbounded(::LinearMap) as follows:

function isbounded(lm::LinearMap)::Bool
    if iszero(lm.M) || isbounded(lm.X)
        return true
    end
    return isbounded_unit_dimensions(lm)
end

Another sufficient criterion is that lm.M is invertible. Since Julia does not provide such a check, we discussed to add a sufficient check of a matrix being invertible/not singular as follows (based on this reference):

function isinvertible_sufficient(M::AbstractMatrix; cond_tol::Number=1e6)
    return cond(M) < cond_tol
end

i think that we can use cond
it correspond to the ratio between the largest and smallest singular value of the matrix
if it is bigger than some given tolerance, then the matrix is singular
there will always be some tolerance because we make the computations in floating point

Can we also use this idea for ExponentialProjectionMap?

[mforets]

I think that it can be use only when the matrix Le^A R is a square matrix.