[flang][cuda] Add missing interface for __cosf (#153306)

`__cosf` is mentioned to be supported here:
https://docs.nvidia.com/hpc-sdk/compilers/cuda-fortran-prog-guide/index.html#fortran-device-modules

Add the missing interface with a bind c name linking it to `__nv_cosf`

Author: clementval

flang/test/Lower/CUDA/cuda-device-proc.cuf

@@ -59,7 +59,8 @@ attributes(global) subroutine devsub()
   res = __brev(ai)
   resl = __brev(al)
   res = __clz(ai)
-  res = __clz(al)
+  res = __clz(al)  
+  af = __cosf(af)
   ai = __mul24(ai, ai)
   ai = __umul24(ai, ai)
 end
@@ -110,11 +111,10 @@ end
 ! CHECK: %{{.*}} = fir.call @__nv_brevll(%{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i64) -> i64
 ! CHECK: %{{.*}} = fir.call @__nv_clz(%{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i32) -> i32
 ! CHECK: %{{.*}} = fir.call @__nv_clzll(%{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i64) -> i32
-
+! CHECK: %{{.*}} = fir.call @__nv_cosf(%{{.*}}) proc_attrs<bind_c> fastmath<contract> : (f32) -> f32
 ! CHECK: %{{.*}} = fir.call @__nv_mul24(%{{.*}}, %{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i32, i32) -> i32
 ! CHECK: %{{.*}} = fir.call @__nv_umul24(%{{.*}}, %{{.*}}) proc_attrs<bind_c> fastmath<contract> : (i32, i32) -> i32
 
-
 subroutine host1()
   integer, device :: a(32)
   integer, device :: ret