add benchmarks ; add explicit #[inline] annotations

This diff adds benchmarks to get more info regarding Issue #1.

It appears that the remaining difference between the "simple"
case and the "cargo" case is the result of a difference in
performance between using `match` and `if` for tight loops.

I suspect it's because of the way that match arms get reordered:
if I manually reorder the "if" statement, I can reproduce the
match performance.

Also added a couple #[inline] annotations in tables.rs, though
the difference in performance in my measurements is negligible.

Bumped version number to 0.1.1.

scripts/unicode.py

@@ -206,6 +206,7 @@ def emit_charwidth_module(f, width_table):
     #[cfg(feature = "no_std")]
     use core::result::Result::{Ok, Err};
 
+    #[inline]
     fn bsearch_range_value_table(c: char, is_cjk: bool, r: &'static [(char, char, u8, u8)]) -> u8 {
         #[cfg(feature = "no_std")]
         use core::cmp::Ordering::{Equal, Less, Greater};
@@ -226,6 +227,7 @@ def emit_charwidth_module(f, width_table):
 """)
 
     f.write("""
+    #[inline]
     pub fn width(c: char, is_cjk: bool) -> Option<usize> {
         match c as usize {
             _c @ 0 => Some(0),          // null is zero width