Adjust the precision of J0/J1 based on new tests

It seems that these results are not target dependent, the high ULP
results can be hit on x86 as well.

Author: tgross35

crates/libm-test/src/precision.rs

@@ -26,10 +26,7 @@ pub fn default_ulp(ctx: &CheckCtx) -> u32 {
         // Overrides that apply to either basis
         // FMA is expected to be infinite precision.
         (_, Id::Fma | Id::Fmaf) => 0,
-        (_, Id::J0 | Id::J0f | Id::J1 | Id::J1f) => {
-            // Results seem very target-dependent
-            if cfg!(target_arch = "x86_64") { 4000 } else { 800_000 }
-        }
+        (_, Id::J0 | Id::J0f | Id::J1 | Id::J1f) => 800_000,
         (_, Id::Jn | Id::Jnf) => 1000,
 
         // Overrides for musl