20260226
  Provide a better error message for attempts to reuse a loopdummy as
a dummy.  Example:  with

	param p{1..3};
	for{i in 1..2}{
		let {i in 1..3} p[i] := Uniform01();
		display p;
		}

in file "foo", the invocation "ampl foo" now gives

		syntax error: attempt to reuse loopdummy as dummy
	context:  let {i in  >>> 1..3} <<<  p[i] := Uniform01();

instead of just

	foo, line 3 (offset 39):
		syntax error
	context:  let {i  >>> in  <<< 1..3} p[i] := Uniform01();

  Fix an obscure bug in determining which variables are used nonlinearly.
Previously
	model hs114.mod
	option solver conopt, linelim 0;
	solve;
gave
	CONOPT 3.17A: variable _svar[10]: Nonlinear variable does not appear in Hessian.
	CONOPT error: Status not called.
(with an older version of conopt).  Now it gives
	CONOPT 3.17A: Locally optimal; objective -1768.806964
	14 iterations; evals: nf = 25, ng = 15, nc = 122, nJ = 15, nH = 0, nHv = 0

New option keep_dv_eqn (default 0).  When $keep_dv_eqn is 1, a defined variable
declaration is separated in a variable declaration and a constraint that the
variable should have the specified value.  For example,

	var x = y^2 + exp(z);

is treated as

	var x;
	s.t. =x: x == y^2 + exp(z);

The name of the constraint "=x" appears in the *.row file if $auxfiles
mentions "r" and appears in solexpand output but cannot be used in
commands.  By experimenting with 0 and 1 values of options
keep_dv_eqn, linelim, and substout, one may find settings that work
well for a particular problem and solver.