Fix nanmean input validation

[sahilphad07-sudo]

nanmean did not correctly handle logical inputs and did not validate some invalid dimension arguments. This change adds logical input support and validates non-finite, duplicate, and empty dimension arguments.

Before

octave:1> nanmean (logical ([0 1 1]))
error: nanmean: X must be numeric.

octave:2> nanmean (ones (3), [])
ans =

   1   1   1
   1   1   1
   1   1   1

For Inf and duplicate dimension cases, the previous implementation did not perform the required validation.

After

octave:1> nanmean (logical ([0 1 1]))
ans = 0.6667

octave:2> nanmean (ones (3), Inf)
error: nanmean: DIM must be a positive integer.

octave:3> nanmean (ones (3), [1 Inf])
error: nanmean: VECDIM must be a vector of positive integers.

octave:4> nanmean (ones (3), [1 1])
error: nanmean: VECDIM must be a vector of positive integers.

octave:5> nanmean (ones (3), [])
error: nanmean: DIM must be a positive integer.

MATLAB Behavior

>> nanmean(logical([0 1 1]))

ans =

    0.6667

>> nanmean(ones(3), Inf)
Error using sum
Dimension argument must be a positive integer scalar, a vector of unique positive integers, or 'all'.

>> nanmean(ones(3), [1 Inf])
Error using sum
Dimension argument must be a positive integer scalar, a vector of unique positive integers, or 'all'.

>> nanmean(ones(3), [1 1])
Error using sum
Elements of a dimension vector must be unique.

>> nanmean(ones(3), [])
Error using mean
Dimension argument must be a positive integer scalar, a vector of unique positive integers, or 'all'.

Regression Tests Added

• Logical input.
• Non-finite scalar dimension.
• Non-finite dimension in a vector.
• Duplicate dimensions.
• Empty dimension argument.

Test Results

• 26 out of 26 nanmean BISTs pass.