CMake updates (#302)

* Add BUILD_SHARED_LIBS option to choose between shared and static lib

* Fix build break for Android aarch64

* Install built target

* Set the correct version number

* Don't add include dirs through CFlags. Include dirs are already set as target property

* Install all relevant headers, not just from include dir

* Update note about BUILD_SHARED_LIBS in README.md

* Fix mistake in README.md, option(BUILD_SHARED_LIBS ON) sets the lib to shared by default

Merge pull request #297 from JuliaMath/dependabot/github_actions/code…

…cov/codecov-action-4

Bump codecov/codecov-action from 3 to 4

Correctly round double precision sqrt (#256)

As discussed in JuliaLang/julia#43786, openlibm's sqrt function is incorrectly rounded for i387. IEEE requires correct rounding for these functions and LLVM relies on it. Fix that by setting the precision in the FPU control word (see e.g. e_ceil.S for similar FPU modifications).

Bump version to 0.8 (#248)

Bump the SOVERSION as well since we have some new exports and such.

Merge pull request #228 from JuliaMath/aa/hypotl

Fix incorrect results in `hypotl` near underflow

Merge pull request #221 from maleadt/tb/static_fenv

Merge pull request #219 from maleadt/tb/dont_export_fenv

Revert "Export `fenv` functions on all platforms (#213)"

Fix Apple Silicon build (#214)

My previous Apple Silicon build went through, so I thought it already
worked, but it turns out it accidentally built an armv7 build instead.
This actually fixes the Apple Silicon build. One thing to note in
particular is that on Apple Silicong `long double` is the same as
`double` while on Linux `long double` is a 128 bit double-double
format.

Co-authored-by: Elliot Saba <[email protected]>

Export `fenv` functions on all platforms (#213)

Win32 has been using a hack to switch the `fenv` functions from `static`
to `DLLEXPORT`, we apply that hack to all platforms that do not use a
host `fenv.h`.

Bump SONAME as discussed in #200