#2515 unreliable scrollbar tests

[Don Porter]

When the scrollbar.test file is
run alone, sometimes these tests
fail:

==== scrollbar-3.36 ScrollbarWidgetCmd procedure, "fraction" option FAILED
==== Contents of test case:

format {%.6g} [.s fraction 4 179]

---- Result was:
0.98773
---- Result should have been (exact matching):
1
==== scrollbar-3.36 FAILED

==== scrollbar-3.38 ScrollbarWidgetCmd procedure, "fraction" option FAILED
==== Contents of test case:

format {%.6g} [.s fraction 4 178]

---- Result was:
0.981595
---- Result should have been (exact matching):
0.993711
==== scrollbar-3.38 FAILED

==== scrollbar-6.12 ScrollbarPosition procedure FAILED
==== Contents of test case:

.s identify 8 19

---- Result was:
trough1
---- Result should have been (exact matching):
arrow1
==== scrollbar-6.12 FAILED

==== scrollbar-6.29 ScrollbarPosition procedure FAILED
==== Contents of test case:

.s identify 8 180

---- Result was:
trough2
---- Result should have been (exact matching):
arrow2
==== scrollbar-6.29 FAILED

==== scrollbar-6.35 ScrollbarPosition procedure FAILED
==== Contents of test case:

.s identify 18 100

---- Result was:

---- Result should have been (exact matching):
trough2
==== scrollbar-6.35 FAILED

I haven't been able to isolate the cause, but
it has something to do with interpretation
of coordinate values relative to the coords
of the scrollbar widget. These are sensitive
to context and/or history in a way that makes
the tests unreliable.