Improve processing of different file formats when generating content

### Background

Now, when processing the files template engine treat all the files (unless specified as `copyOnly`) as:
- textual
- trying to perform replace and other operations
This causes performance issues and accidentally errors when generating the content when the content of the file matches the replacement.

### Suggestion
Improve detection of the file format and the rules upon that.
Not limited to:
- enrich the defaults for "copy-only", for example in web templates libs are often included, etc
- if the file is detected binary, do "copy-only" mode by default
- if the file is well-known textual type, process operations in it

Ways for detection:
- probe first bytes to detect binary template
- https://www.nuget.org/packages/EmptyFiles

Consider same rules in template testing framework for scrubbing.