Fixed parsing of float literals.

Float literals were not parsed properly when their fractional part
exceeded 53 significant bits. This affected in particular math.PI
and math.E. Rather than reinventing the wheel, this patch reuses
C's strtod() implementation, which already does the heavy lifting
with respect to correctness, though some caution is necessary to
keep float parsing locale-independent.

tests/float/tfloat4.nim

@@ -0,0 +1,42 @@
+import math, strutils
+
+proc c_sprintf(buf, fmt: cstring) {.importc:"sprintf", header: "<stdio.h>", varargs.}
+
+proc floatToStr(f: float64): string =
+  var buffer: array[128, char]
+  c_sprintf(buffer, "%.16e", f)
+  result = ""
+  for ch in buffer:
+    if ch == '\0':
+      return
+    add(result, ch)
+
+let testFloats = [
+  "0", "-1", "1", "1.", ".3", "3.3", "-.3", "-99.99",
+  "1.1e10", "-2e100", "1.234e-10", "1.234e+10",
+  "-inf", "inf", "+inf",
+  "3.14159265358979323846264338327950288",
+  "1.57079632679489661923132169163975144",
+  "0.785398163397448309615660845819875721",
+  "1.41421356237309504880168872420969808",
+  "0.707106781186547524400844362104849039",
+  "2.71828182845904523536028747135266250",
+  "0.00097656250000000021684043449710088680149056017398834228515625"
+]
+
+for num in testFloats:
+  assert num.parseFloat.floatToStr.parseFloat == num.parseFloat
+
+assert "0".parseFloat == 0.0
+assert "-.1".parseFloat == -0.1
+assert "2.5e1".parseFloat == 25.0
+assert "1e10".parseFloat == 10_000_000_000.0
+assert "0.000_005".parseFloat == 5.000_000e-6
+assert "1.234_567e+2".parseFloat == 123.4567
+assert "1e1_00".parseFloat == "1e100".parseFloat
+assert "3.1415926535897932384626433".parseFloat ==
+       3.1415926535897932384626433
+assert "2.71828182845904523536028747".parseFloat ==
+       2.71828182845904523536028747
+assert 0.00097656250000000021684043449710088680149056017398834228515625 ==
+     "0.00097656250000000021684043449710088680149056017398834228515625".parseFloat