Added tests for "Using C++ in Cython" part 4

Added tests for "Using C++ in Cython" part 5

Added tests for "Using C++ in Cython" part 7

Also exclude the parser from compilation, since its binary size is almost 1MB, but it only contributes a couple of percent to the overall runtime.

Remove some large binary modules from the compiled Cython installation that contribute almost nothing to the compiler speedup (and thus only waste disk space and build time).

docs: Added a note about the performance of slicing C++ stl containers.

Fix test_itimer on x32

On x32, tv_sec and tv_usec are in fact long long rather than the standard long,
and so in Python 2.x are automatically converted to Python longs, causing the
test to fail with:

    Failed example:
        test_itimer(10, 2)
    Expected:
        (10, 2)
    Got:
        (10L, 2L)

To fix this, we should explicitly call the int constructor.

Add safety checks to prevent exception subtype checks with non-types. This raises a TypeError in Py3 now and we should handle that at some point.

Closes #2425.

Refactoring of the extend function in "using C libraries" to make it simpler.

Added tests to "Using Parallelism" part 4

Added tests to "Using Parallelism" part 3

Added tests to "Extension types" part 2

Added tests to "Extension types" part 3

Added tests to "Early binding for speed" part 2

Added tests to "Implementing the buffer protocol" part 2

Added tests to "Implementing the buffer protocol" part 3

Added tests to "Using Parallelism" part 2

Adding tests for "Unicode and passing strings" part 13

Adding tests for "Unicode and passing strings" part 14

Adding tests for "Unicode and passing strings" part 12