Upgrade blueprint benchmark template

[stephannv]

Hey, I was working on Blueprint perf improvements today and I got improved numbers. It's not close to to_html times but at least it's something.
Two things that contributed:

• Using String::Builder instead IO::Memory 🤦🏽
• Allowing passing content without using blocks, eg. h1("Hello").

         ecr   3.48M (286.99ns) (± 1.11%)  4.27kB/op        fastest
     to_html   1.90M (525.41ns) (± 0.88%)  5.52kB/op   1.83× slower
   blueprint 481.79k (  2.08µs) (± 0.78%)   9.8kB/op   7.23× slower
html_builder 118.20k (  8.46µs) (± 0.85%)  10.4kB/op  29.48× slower
       water 114.23k (  8.75µs) (± 0.59%)  11.2kB/op  30.50× slower
     markout  89.04k ( 11.23µs) (± 0.74%)  15.6kB/op  39.13× slower

[stephannv]

I realized why the speed improvement. I forgot to escape content passed via argument without using block 😅
After the fix (stephannv/blueprint#55), it returned to the ~9x slower level. Closing here.

[sbsoftware]

Too bad it didn't work out!

I'm was also considering to support providing content via arguments, but I doubt it would bring any performance benefit to to_html.cr as blocks are eliminated/flattened when compiling anyway. Do you think it is worth the effort usability-wise?

[stephannv]

@sbsoftware At first I didn't like, but I started to enjoy using on situations like:

h1 "Title here"

span "More content here"

footer @my_var

It feels a little like slim/pug template languages

Ps. I worked on some improvements today and I could improve the numbers a little bit, I will open a new PR since I force pushed my branch and I can't reopen this one.