Fix undefined behavior in shifting

[escherstair]

Rule 6.5.7 of ISO/IEC 9899:1999 (E) standard describes Bitwise shift operators.
Paragraph 5 states:

The result of E1 >> E2 is E1 right-shifted E2 bit positions. If E1 has an unsigned type
or if E1 has a signed type and a nonnegative value, the value of the result is the integral
part of the quotient of E1 divided by the quantity, 2 raised to the power E2. If E1 has a
signed type and a negative value, the resulting value is implementation-defined.

So, right shifing a negative number is implementation-defined (undefined behavior), if it is really negative, basically because it depends on how the sign bit is implemented.
But nobody can know if during the execution E1 will be negative or positive.
This is one of the reasons why shifting a signed integer (that can be negative, during the program execution) is not a good idea.
Shinfting unsigned integeres only stays on the safe side for every scenario.

Here are some points in the source code where signed integers are shifted:

• CMSIS-DSP/Include/dsp/none.h

  Line 139 in 93c4e4a

  return ((q31_t) (x >> 32) != ((q31_t) x >> 31)) ?

• CMSIS-DSP/Include/dsp/none.h

  Line 149 in 93c4e4a

  return ((q31_t) (x >> 32) != ((q31_t) x >> 31)) ?

• CMSIS-DSP/Include/dsp/none.h

  Line 160 in 93c4e4a

  ((0x7F ^ ((q7_t) (x >> 31)))) : (q7_t) x;

• CMSIS-DSP/Include/dsp/none.h

  Line 170 in 93c4e4a

  ((0x7FFF ^ ((q15_t) (x >> 31)))) : (q15_t) x;

Is it possible having them fixed so that tools for static analysis do not complain about them?
Thanks