extra tests, plus fill value fixes for string type

[edhartnett]

This contains extra tests, and fixes for the fill values of strings.

Also contains the rename tests, including a couple of commented out lines which show where some rename problems remain.

Fixes #727.
Fixes #732.

[edhartnett]

The failure we are seeing is this:

The following tests FAILED:
	127 - nc_test4_tst_hdf4_read_var (Failed)

Yet this test does not include hdf4 as far as I can tell, so I don't know what is going on.

I suspect this has something to do with tst_interops2 and parallel builds. I'm not sure that the CMakeLists.txt is correct in how tst_interops2. It has the following:

IF(USE_HDF4_FILE_TESTS AND NOT MSVC)
    add_bin_test_no_prefix(tst_interops2)
    build_bin_test_no_prefix(tst_interops3)
    add_sh_test(nc_test4 run_get_hdf4_files)
    add_bin_test(nc_test4 tst_bug324)
    add_sh_test(nc_test4 tst_formatx_hdf4)
    build_bin_test(tst_chunk_hdf4)
    add_sh_test(nc_test4 run_chunk_hdf4)
    add_bin_test(nc_test4 tst_h4_lendian)
    add_sh_test(nc_test4 tst_hdf4_read_var)
    SET_TESTS_PROPERTIES(nc_test4_tst_hdf4_read_var PROPERTIES DEPENDS tst_interops2)
ENDIF()

tst_interops2 must run before tst_hdf4_read_var.sh. The SET_TESTS_PROPERTIES should assure this.

For other C tests that are run from shell scripts, we stopped trying to track dependencies between the running of the C program and the shell script. I instead changed the shell scripts to run the C programs directly, and then the build system just has to make sure the C program is built before the shell script is run. This turns out to be a lot simpler and easier.

I'm not sure what is going on with tst_interops2. I think we should probably do the same with it as we do with the other C test programs that are checked by shell script.

[edhartnett]

I retriggered a build, but a bunch of them timed out. I will leave this and let Ward re-run tests when it is near merging.