Redesign and set up restrictions for simplication of expressions

[Peter Aronsson]

Simplification of expressions can lead to numerical problems. For instance a/b/c can if rewritten to a/(b*c) lead to overflow and loss of precision.

There is therefore a need to set up some rules and restrictions on what simplifications can be allowed. Also, the simplifier should probably not succeed and change an expression if it is not "more simple". The reason being that the modeller probably knows best and if he writes a/b/c instead of a/(b*c) he might have a good reason for it.

[Adrian Pop]

This should be part of milestone Red October (deadline 2010-10-15).

[Francesco Casella]

See #4293