- Return HTML directly, drop rmarkdown
- Improve classes and specificity of rules for result styling
- Improve font size of full page regex results

Replaced a few calls to stringr with stringi (which has to stay for unicode encoding function). Not likely that people won't have stringr, but makes the dependencies a little better.

- Fixed min-height for inputs and controls
- Inputs, controls and outputs won't overlap
- Build out HTML instead of using fillCol()
- Modified textInputCode() to allow additional items into input-group for library button
- Use input-group instead of form-group for textInputCode()

Thanks @GegznaV -- As these are three separate regular expressions (`\W` `\D` `\S`), so I suggest adding "not" before each explanation to stress their meaning.

Also adds regexplain_web()

Track # of result groups in regex result data frame as idx$pass. Refactored span collating.

Global matching implemented inside run_regex(), nothing above initial regex matching works now.

- Removed knitr parameter
- Add result_only (not in knitr)
- Soft deprecate kntir parameter
- Separate group.css from other css styles

Fixes issues with adding <span> and </span> around zero-length groups and more generally groups that start and end at the same index.

Ensures that g00 wraps all others, then at index with close/opens, closes open groups first before opening new groups.

and restore when input$pattern is empty