ao: simplify hack for float atomics

stdatomic.h defines no atomic_float typedef. We can't just use _Atomic
unconditionally, because we support compilers without C11 atomics. So
just create a custom atomic_float typedef in the wrapper, which uses
_Atomic in the C11 code path.

audio/out/ao.c

@@ -642,19 +642,7 @@ void ao_print_devices(struct mpv_global *global, struct mp_log *log)
 
 void ao_set_gain(struct ao *ao, float gain)
 {
-    uint_least32_t v = 0;
+    atomic_store(&ao->gain, gain);
-    assert(sizeof(gain) <= sizeof(v));
-    memcpy(&v, &gain, sizeof(gain));
-    atomic_store(&ao->gain_fi, v);
-}
-
-static float ao_get_gain(struct ao *ao)
-{
-    uint_least32_t v = atomic_load_explicit(&ao->gain_fi, memory_order_relaxed);
-    float gain;
-    assert(sizeof(gain) <= sizeof(v));
-    memcpy(&gain, &v, sizeof(gain));
-    return gain;
 }
 
 #define MUL_GAIN_i(d, num_samples, gain, low, center, high)                     \
@@ -669,7 +657,7 @@ static float ao_get_gain(struct ao *ao)
 
 static void process_plane(struct ao *ao, void *data, int num_samples)
 {
-    float gain = ao_get_gain(ao);
+    float gain = atomic_load_explicit(&ao->gain, memory_order_relaxed);
     int format = af_fmt_from_planar(ao->format);
     if (gain == 1.0f)
         return;

audio/out/internal.h

@@ -70,8 +70,8 @@ struct ao {
     // Internal events (use ao_request_reload(), ao_hotplug_event())
     atomic_int events_;
 
-    // Float gain multiplicator, reinterpret-casted to int.
+    // Float gain multiplicator
-    atomic_uint_least32_t gain_fi;
+    mp_atomic_float gain;
 
     int buffer;
     double def_buffer;

osdep/atomic.h

@@ -24,6 +24,7 @@
 
 #if HAVE_STDATOMIC
 #include <stdatomic.h>
+typedef _Atomic float mp_atomic_float;
 #else
 
 // Emulate the parts of C11 stdatomic.h needed by mpv.
@@ -36,6 +37,8 @@ typedef struct { long long v;          } atomic_llong;
 typedef struct { uint_least32_t v;     } atomic_uint_least32_t;
 typedef struct { unsigned long long v; } atomic_ullong;
 
+typedef struct { float v;              } mp_atomic_float;
+
 #define ATOMIC_VAR_INIT(x) \
     {.v = (x)}