bad behavior (bug): substitute using a matrix of values

substituting using a matrix (not Options, not a ring map) can still give the wrong/unexpected answer. Here are two situations (one over QQ, the other over CC_53).

It should be easy to fix this, Dan, as the "Options" version was changed to produce these correct values recently.

```
R = QQ[x]
m = matrix{{1/2 * x}}
sub(m, {x=>1_ZZ}) -- is correct behavior
sub(m, matrix{{1}}) -- just wrong. Should give behavior on previous line
f = map(ZZ, R, {1}) -- REALLY: should give an error: cannot construct ring map from R --> ZZ.
f m -- wrong behavior currently. But if f cannot be constructed, this will not be a problem
g = map(QQ[y], R, {1}) -- this one works now, and is the behavior we want
g m

R = CC[x]
m = matrix{{ii * x}}
sub(m, {x=>1_ZZ}) -- is correct behavior
sub(m, matrix{{1}}) -- just wrong. Should give behavior on previous line
f = map(RR_53, R, {1}) -- REALLY: should give an error.
f m -- crashes in linalg branch, gives 0 in 1.6. Both are wrong...!
g = map(CC_53[y], R, {1}) -- this one works now, and is the behavior we want
g m
```

Reproduce the two matrix-substitution examples over QQ and CC_53, comparing substitution with the working Options and ring-map cases. Trace the substitution and ring-map behavior involved in those examples; done means matrix substitution matches the scalar substitution and invalid maps produce errors rather than incorrect values or crashes.

Written by the indexing model from the issue text.