[CppAD] User Atomic Functions

[Brad Bell]

The old user atomic function interface
     http://www.coin-or.org/CppAD/Doc/old_atomic.xml
has been deprecated. The new interface
     http://www.coin-or.org/CppAD/Doc/atomic_base.xml
should be easier to use and faster.

In addition, atomic_base has been used to create an easy to use and 
flexible method for checkpointing functions for which the operation 
sequence does not depend on the arguments; see;
     http://www.coin-or.org/CppAD/Doc/checkpoint.xml
     http://www.coin-or.org/CppAD/Doc/glossary.xml#Operation.Sequence

One use of the atomic_base feature could be to extend CppAD to include 
atomic matrix operations. A matrix multiply operation has been included 
as an example; see
     http://www.coin-or.org/CppAD/Doc/atomic_mat_mul.cpp.xml

If one made solving linear equations an atomic operation, one would not 
need to retape the operations sequence when the choice of piviot rows or 
columns changed; i.e., the operation sequence would not depend on the 
argument values. This type of extension of the atomic operations could 
be used for any algorithm, not just matrix operations.