after updating all of the movXYZ functions i noticed a few build in an "Endpoints", # option for efficiency vs 'using default shrink. This hits an error if non-uniform samplepoints are defined as the two aren't compatible. added a check to avoid that. pushed to default as

https://hg.savannah.gnu.org/hgweb/octave/rev/a70dacb39639

revision of movfun pushed as
https://hg.savannah.gnu.org/hgweb/octave/rev/8b0099858d35

that replaces the inputParser checks with manual input checking and error messages. previously i had added a fixme noting that the various inputparser elements were consuming ~25-30% of runtime. This is likely from inputParser being m-code as well as the input validation checks being done as anonymous functions, which are known to add nontrivial overhead.

Additionally, the input validation error messages from inputParser are rather uninformative. when choosing between:

>> movfun (@sum, 1:10, 3, "Endpoints", "foo")
error: movfun: failed validation of 'Endpoints'.  Checked with "@(x) any (strcmpi (x, valid_bc)) || (isnumeric (x) && isscalar (x))"

and

>> movfun (@sum, 1:10, 3, "Endpoints", "b")
error: movfun: ENDPOINTS must be a numeric scalar or a valid Endpoint method
error: called from
    movfun at line 227 column 13

I think the latter is preferable.  If there's a way to get better messaging directly from inputParser i haven't found it other than writing separate, individual, subfunctions to use for the checks, and at that point writing the faster parsing isn't much harder.

Samplepoints option implementation in movslice and movfun pushed to default as https://hg.savannah.gnu.org/hgweb/octave/rev/b541ead8c01c

The nanflag option would be best implemented by passing that option through to the underlying builtin functions. I believe all of the MATLAB base functions have that option, while they're still pending on many of ours.  I'm going to move that option to a separate report for tracking.

Marking this one as ready for test.