redo-whichdo: fix a bug where the last dir was checked twice, and add tests.

When we can't find a .do file, we walk all the way back to the root
directory.  When that happens, the root directory is actually searched
twice.  This is harmless (since a .do file doesn't exist there anyway)
but causes redo-whichdo to produce the wrong output.

Also, add a test, which I forgot to do when writing whichdo in the
first place.

To make the test work from the root directory, we need a way to
initialize redo without actually creating a .redo directory.  Add a
init_no_state() function for that purpose, and split the necessary path
functions into their own module so we can avoid importing builder.py.

t/260-whichdo/exists.do

@@ -0,0 +1,25 @@
+exec >&2
+
+a=$(cd fakesub && redo-whichdo ../a/b/x.y.z)
+# if sh doesn't abort after the above, then it found a .do file as expected
+
+# Note: we expect redo-whichdo to return paths relative to $PWD at the time
+# it's run, which in this case is fakesub.
+# Likely bugs would be to return paths relative to the start dir, the .redo
+# dir, the current target dir, the requested target dir, etc.
+b=$(cat <<EOF
+../a/b/x.y.z.do
+../a/b/default.y.z.do
+../a/b/default.z.do
+../a/b/default.do
+../a/default.y.z.do
+../a/default.z.do
+../a/default.do
+../default.y.z.do
+EOF
+)
+
+if [ "$a" != "$b" ]; then
+    printf 'redo-whichdo mismatch.\n\ngot:\n%s\n\nexpected:\n%s\n' "$a" "$b"
+    exit 11
+fi