sw: Add optimized AXPY, Covariance, SYRK and Doitgen kernels

[colluca]

• Allow $(APP)_INCDIRS to be extended on a per-app basis.
• Extend axpy kernel
  • Support multiple tiles and alternative implementations
  • Optimize TCDM conflicts by appropriately placing data in TCDM
  • Support double-buffering
• Extend covariance kernel
  • Support multiple tiles and alternative implementations
  • Remove divisions by statically generating reciprocals 1 / n and 1 / (n - 1)
  • Add optimized and baseline implementations
  • Support double-buffering
• Add syrk and doitgen kernels (double-buffered, optimized).
• Extend clean-traces target to remove all artifacts of make traces.
• Remove .gitignore and Makefile files, outdated after sw: Switch to non-recursive Make build system #171.
• Do not exit on exception in gen_trace.py, so performance metrics are dumped.
• Uniformize kernels to generate data values in range [-1.0, 1.0] to avoid optimistic power results from using integers.
• Increase error threshold on GEMM kernel, as opposite sign operands can lead to cancellation and increased errors.