FMSLogo Code

Remove settree__tree().
This had an unstated side effect of setting the treepair's generation.
Worse, the generation had to be immediately modified by all callers, which convinced me that it'd would be better for the callers to construct the treepair.

Cleanup: Remove unneeded local variable "s" in cnv_node_to_numnode.

Cleanup: Remove logic to skip over the leading + in cnv_node_to_numnode.
This was dead code because numberp() doesn't allow numbers to begin with a +.

Cleanup: Remove some unneeded vref/deref calls in intern_primitive().
It looked like these calls canceled each other out.

Refactoring: rework intern_primitive() to reduce duplicated code.

Fix some misinformation in a comment.
ufun != NIL doesn't mean "not a lambda" it means "user-defined procedure".

Cleanup: Remove a block of code that indent to the tracelevel for singlestepping.

In UCBLogo, the code looks like:

if (ufun != NIL && flag__caseobj(ufun, PROC_STEPPED)) {
if (tracing) {
int i = 1;
while (i++ < trace_level) print_space(stdout);
}
print_node(stdout, this_line);
(void)reader(stdin, " >>> ");
}

This makes sense in UCBLogo because the trace output is interwoven with the single-step dialog.
However, in FMSLogo, the single-step dialog is a separate dialog box, so adding
indentation to the commander history, where the trace output goes, doesn't make sense.

Refactoring: Extract indent_to_trace_level() to reduce duplicated code.

Refactoring: Extract function retreeify_running_proc()
This is a chunk of logic from the evaluator that doesn't access much of the evaluator's state.
Moving this logic into its own function enables getting rid of a label in evaluator().

Correct some misinformation in a comment.
When the Unbound generation is propagated up, it's not applied to the first line.
It's applied to the "bodylist", which is the container for the first node.