feat: use more precise pi

with pretty_eval, I still get:
-9.194801691 * 10^-618

but at least that's more accurate than the prior:
1.224646799 * 10^-16

src/nodes/parse.rs

@@ -1,7 +1,10 @@
 use core::f64;
 use std::str::FromStr;
 
-use rug::{Float, float::ParseFloatError};
+use rug::{
+    Float,
+    float::{self, ParseFloatError},
+};
 
 use crate::nodes::{Function, Node, Nodes, Operator};
 
@@ -10,7 +13,7 @@ impl Node {
         Ok(match s {
             "c" => Self::from(299_792_458),
             "A" => Self::from(6.02214076e23),
-            "pi" | "π" => Self::from(f64::consts::PI),
+            "pi" | "π" => Self::from(float::Constant::Pi),
             // Planck's constant, in J*s (joule-seconds)
             "h" => Self::from(6.62607015e-34),
             "e" => Self::from(f64::consts::E),

src/nodes/tests.rs

@@ -51,3 +51,8 @@ fn constant_implicit_mult() {
     assert_eq!(eval("2pi"), 6.283185307179586);
     assert_eq!(eval("pi2"), 6.283185307179586);
 }
+
+#[test]
+fn pi_precision() {
+    assert_eq!(eval("sin(pi)"), 0.)
+}