Html view rendering performance

[dv00d00]

I've added a stripped implementation of html view rendering to the techempower tests repo. It finished test run today. These are results https://www.techempower.com/benchmarks/#section=test&runid=3da523ee-fff1-45d8-9044-7feb532bf9ee&hw=ph&test=fortune&w=zik0zj-hr4e7z&a=2

name	rps
giraffe-stripped	174,979
giraffe	41,136

While testing stripped version I observed that to render fortunes html the html generation itself produces around 100 KB allocation.

What I did is very straightforward, I just rendered html to the memory stream instead of relying on string concatenation as in current implementation, it surely can be optimized more. @gerardtoconnor has investigated this area deeply. Please take a look at this https://github.com/gerardtoconnor/ViewEngineTests/blob/master/TemplateViewEngine.fs

Fortunes tests were added in this PR TechEmpower/FrameworkBenchmarks#3863

[gerardtoconnor]

Great work @dv00d00 !! I think your solution it great, in line with my ByteViewEngine & already very high up there in the rankings.

I'm looking into how we can juice more performance out of the framework, once that is done I can look into the templating a bit more but I think you're 90% there.

[dustinmoris]

First of all congrats on these great results! Secondly I would love to get your changes from string concatenation to memory stream put into the current GiraffeViewEngine. A PR is more than welcome! Let me know if this is something you would like to work on or if it is up for grabs for somebody else!

Thanks a lot!

[dustinmoris]

Release is out: https://github.com/giraffe-fsharp/Giraffe/releases/tag/v3.0.0

Thanks for the great contributions! :) 👍