[MERGE] rand_chars() optimization

[Martin Pool]

On 3/14/07, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> This patch, like most optimizations, reduces legibility.  So far, I
> haven't seen any evidence that it makes any Bazaar operation observably
> faster.  So to my mind, there's no justification for including it.

Yes, I'd like to see some evidence of an overall speedup.  I guess the
most likely place would be in adding many new files.

-- 
Martin