Reorder parsed_number_string_t fields to avoid padding (fixes #418)

Placing the parse_error field next to the booleans shrinks the struct
from 72 to 64 bytes on 64-bit systems, recovering a large performance
regression observed with MSVC. Add a static_assert to guard the size.

include/fast_float/ascii_number.h

diff --git a/include/fast_float/ascii_number.h b/include/fast_float/ascii_number.h
index e814f12..4d68eaa 100644
--- a/include/fast_float/ascii_number.h
+++ b/include/fast_float/ascii_number.h
@@ -308,13 +308,16 @@
   int64_t exponent{0};
   uint64_t mantissa{0};
   UC const *lastmatch{nullptr};
+  // The field order matters: placing 'error' next to the booleans avoids
+  // padding, keeping the struct at 64 bytes on 64-bit systems instead of 72.
+  // See https://github.com/fastfloat/fast_float/issues/418
+  parse_error error{parse_error::no_error};
   bool negative{false};
   bool valid{false};
   bool too_many_digits{false};
   // contains the range of the significant digits
   span<UC const> integer{};  // non-nullable
   span<UC const> fraction{}; // nullable
-  parse_error error{parse_error::no_error};
 };
 
 using byte_span = span<char const>;

tests/basictest.cpp

diff --git a/tests/basictest.cpp b/tests/basictest.cpp
index 9b45822..cde42bc 100644
--- a/tests/basictest.cpp
+++ b/tests/basictest.cpp
@@ -84,6 +84,12 @@
   }
 }
 
+// Guard against padding regressions that hurt performance.
+// See https://github.com/fastfloat/fast_float/issues/418
+static_assert(sizeof(void *) != 8 ||
+                  sizeof(fast_float::parsed_number_string_t<char>) <= 64,
+              "parsed_number_string_t should fit in 64 bytes");
+
 #define FASTFLOAT_STR(x) #x
 #define SHOW_DEFINE(x) printf("%s='%s'\n", #x, FASTFLOAT_STR(x))