riscv: Fix feholdexcept()

[AlekseyZhmulin]

The feholdexcept() function must store the current floating point environment in *__envp.
Related to #321

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 39.03%. Comparing base (3cfad14) to head (188b8f4).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #324   +/-   ##
=======================================
  Coverage   39.03%   39.03%           
=======================================
  Files         233      233           
  Lines        6151     6151           
  Branches     1608     1608           
=======================================
  Hits         2401     2401           
  Misses       3397     3397           
  Partials      353      353           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[inkydragon]

We're doing the same thing as BSD.
I'm not sure this works.

https://github.com/freebsd/freebsd-src/blob/master/lib/msun/riscv/fenv.h#L188-L195

[AlekseyZhmulin]

The problem is that feholdexcept() does not save floating point environment, but feupdateenv() loads floating point environment. Therefore, in the lrint() function, "fenv_t env" is not initialized and a random value from the stack is loaded into the fcsr register. https://github.com/JuliaMath/openlibm/blob/master/src/s_lrint.c#L59. This throws an "Illegal instruction" exception.