_valgrind_ rebased to version 3.13.0
The _valgrind_ package has been upgraded to version 3.13.0, which provides a number of bug fixes and enhancements over the previous version. Notable changes are:
* *Valgrind* has been extended in several ways to run large programs. The amount of memory usable by *Valgrind* has been increased to 128 GB. As a consequence, the *Memcheck* tool supports running applications that allocate up to approximately 60 GB. Additionally, *Valgrind* can now load executable files up to 1200 MB in size.
* The tools *Memcheck*, *Helgrind*, and *Massif* can now use a new execution tree (xtree) representation to report heap consumption of the analyzed applications.
* The symbol demangler has been updated to support the C++11 standard and the Rust programming language.
* Failures with long blocks of code using AVX2 instructions on the Intel and AMD 64-bit architecture have been fixed.
* The 64-bit `timebase` register of the PowerPC architecture is no longer modeled by *Valgrind* as only 32-bit.
* Support for the IBM Power Systems architecture has been extended to include the ISA 3.0B specification.
* An alternative implementation of Load-Linked and Store-Conditional instructions for the 64-bit ARM architecture has been added. The alternative implementation is enabled automatically when required. To enable it manually, use the `--sim-hints=fallback-llsc` option.