Fix potential bad memory access at polyroots function

[MiraiHattori]

This PR fixes the attached compiler warning. Potentially if the input of those functions are per-joint linearly changing trajectories, then there may be segmentation fault.

Details

• when the template parameter D=1, then this roots[1] is out of bound.
  

  openrave/include/openrave/mathextra.h

  Line 1233 in 7fdb0b5

  roots[1] = complex<IKReal>(0.4,0.9); // any complex number not a root of unity works

• caller with D=1. if the system accidentally hits it, segmentation fault will happen
  

  openrave/include/openrave/mathextra.h

  Lines 1467 to 1470 in 7fdb0b5

  	}
  	else if( tempcoeffs[4] != 0 ) {
  	polyroots<T, 1>(&tempcoeffs[4], rawroots, numroots);
  	}

[MiraiHattori]

I added -Werror=array-bounds compiler options -> later reverted

[MiraiHattori]

bumped "minor" version (instead of patch version) / added changelog.

[MiraiHattori]

I added -Werror=array-bounds compiler options

compilation failed, taking a look.

[cielavenir]

Could be https://qiita.com/cielavenir/items/3e52823a8193b9fb6989

[MiraiHattori]

Could be https://qiita.com/cielavenir/items/3e52823a8193b9fb6989

With an older boost version, compiler did not catch the false positive out of bounds.

I think it's okay without -Werror=array-bounds in this PR, so reverted.

[MiraiHattori]

@yoshikikanemoto could you review this MR?

[yoshikikanemoto]

@MiraiHattori why is minor version bump? looks like there are no ABI breaking changes? do you have unit tests that produce out-of-bounds access?

[MiraiHattori]

@MiraiHattori why is minor version bump? looks like there are no ABI breaking changes? do you have unit tests that produce out-of-bounds access?

I reverted to patch version bump since there was no interface changes.

I don't have a unit test to reproduce the issue so far, let me make one.