Remove support for long double

[antoinefalisse]

This PR aims to remove support for long double ( #596 ). I have removed pieces of code where long doubles were used and adjusted comments where appropriate. I am not sure that long doubles had to be removed everywhere though. For instance, in conjugate.h, I removed definitions of operators involving long doubles but this may not be the right approach. I did not touch the definitions of the SimTK_NICETYPENAME_LITERAL involving long doubles in common.h (I did not touch TestNiceTypeName.cpp either). Please let me know if this is the intended approach or if I should proceed differently.

Reviewers: @sherm1 and @chrisdembia

---

This change is 

[chrisdembia]

@sherm1, here are the places where long double remains:

SimTKcommon/include/SimTKcommon/Constants.h
37:values in long double precision.
51:macros in long double precision. By using very high precision we ensure 
52:sufficient accuracy for any IEEE long double precision implementation (they 
120:written here in maximal (long double) precision. (These values were generated 
124:the compiler will properly calculate a long double result with no runtime cost.

SimTKcommon/include/SimTKcommon/internal/common.h
918:SimTK_NICETYPENAME_LITERAL(long double);
922:SimTK_NICETYPENAME_LITERAL(std::complex<long double>); 

SimTKcommon/Mechanics/include/SimTKcommon/Orientation.h
57:// long double results which you can cast to smaller sizes if you want.

SimTKcommon/tests/BNTTest.cpp
335:        (long double)NTraits<float>::getEps(), 
336:        (long double)NTraits<double>::getEps());
339:        (long double)NTraits<float>::getSqrtEps(), 
340:        (long double)NTraits<double>::getSqrtEps());
343:        (long double)NTraits<float>::getSignificant(), 
344:        (long double)NTraits<double>::getSignificant());
347:        (long double)NTraits<float>::getTiny(), 
348:        (long double)NTraits<double>::getTiny());

SimTKcommon/tests/TestNiceTypeName.cpp
186:    SimTK_TEST(NiceTypeName<long double>::namestr() == "long double");
191:    SimTK_TEST(NiceTypeName<std::complex<long double>>::namestr() 
192:               == "std::complex<long double>");
206:    SimTK_TEST(NiceTypeName<std::complex<long double>>::xmlstr() 
207:               == "std::complex{long double}");

SimTKmath/Integrators/src/CPodes/sundials/include/sundials/sundials_types.h
19: * float, double or long double.
25: * The legal types for realtype are float, double and long double.
35: * expands to 1.0F. If realtype is defined as a long double,
78:typedef long double realtype;

SimTKmath/Integrators/src/CPodes/sundials/INSTALL_NOTES.txt
283:  (float C-type) if ARG=single, or as a long double C-type if ARG=extended.

SimTKmath/Integrators/src/CPodes/sundials/tests/pend.c
39:  printf("long double  %d\n",(int)sizeof(long double));

I agree with leaving in SimTK_NICETYPENAME_LITERAL(long double) but @sherm1 will have to give input about conjugate<long double>.

I looked through all the changes in the PR and they look good to me.

---

Comments from Reviewable

[sherm1]

This is great! Thanks @antoinefalisse. It is really nice to see all that unneeded code go away. It really cuts down on the complexity and future maintenance headache.

One minor request in BNTTest.cpp, otherwise

+@sherm1 +@chrisdembia (this syntax tells Reviewable to require lgtm's from us - you can use it to request reviewers)

---

Reviewed 21 of 21 files at r1.
Review status: all files reviewed at latest revision, 1 unresolved discussion.

---

SimTKcommon/tests/BNTTest.cpp, line 348 at r1 (raw file):

    printf("Tiny f=%.16Lg d=%.16Lg\n",
        (long double)NTraits<float>::getTiny(), 
        (long double)NTraits<double>::getTiny());

BTW Please change these to (double) and remove the "L" from "Lg" in the four printf format strings - no need for long double anymore.

---

Comments from Reviewable

[sherm1]

P.S. @chrisdembia -- thanks for the long double report . Very helpful!

---

Review status: all files reviewed at latest revision, 1 unresolved discussion.

---

Comments from Reviewable

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@1fb07d6). Click here to learn what that means.
The diff coverage is n/a.

@@            Coverage Diff            @@
##             master     #597   +/-   ##
=========================================
  Coverage          ?   49.85%           
=========================================
  Files             ?      557           
  Lines             ?    70076           
  Branches          ?        0           
=========================================
  Hits              ?    34939           
  Misses            ?    35137           
  Partials          ?        0

Impacted Files	Coverage Δ
SimTKcommon/Scalar/include/SimTKcommon/Scalar.h	100% <ø> (ø)
SimTKcommon/include/SimTKcommon/internal/String.h	2.43% <ø> (ø)
...mon/SmallMatrix/include/SimTKcommon/internal/Vec.h	10.84% <ø> (ø)
...Matrix/include/SimTKcommon/internal/MatrixHelper.h	30% <ø> (ø)
...mon/SmallMatrix/include/SimTKcommon/internal/Mat.h	8.3% <ø> (ø)
...on/Scalar/include/SimTKcommon/internal/conjugate.h	0% <ø> (ø)
...mmon/Scalar/include/SimTKcommon/internal/NTraits.h	14.75% <ø> (ø)
SimTKcommon/include/SimTKcommon/internal/common.h	50% <ø> (ø)
SimTKcommon/Polynomial/src/cpoly.cpp	88.07% <ø> (ø)
SimTKcommon/Polynomial/src/rpoly.cpp	90.41% <ø> (ø)
... and 7 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1fb07d6...c2b20b9. Read the comment docs.

[antoinefalisse]

@sherm1, @chrisdembia, I have made the changes in BNTTest so this PR should now be ready. Thanks!

---

Review status: 20 of 21 files reviewed at latest revision, 1 unresolved discussion.

---

Comments from Reviewable

[sherm1]

Thanks, @antoinefalisse!

---

Reviewed 1 of 1 files at r2.
Review status: all files reviewed at latest revision, all discussions resolved.

---

Comments from Reviewable