Refactor for exceptions

[jim-parry]

Enhancement Project: replace usage of show_error() with exceptions. A class or function, instead of calling show_error to display an error message and then exit, should instead throw a PHP exception.

Benefits: improved error handling, better testability.

Related: enhance or replace error-handling writeup in the user guide.

Procedure:

• In general, replace every show_error() call with a throw new Exception(), or as appropriate, one of the other predefined exceptions.
• Translations would be mostly dropped in that process as they don't make sense for exceptions ... cases where a translation may be necessary will be discussed/decided on the fly.
• And just to make the transition smoother, show_error() itself should be altered to throw Exception.

Process:

• Tackle the affected classes one at a time, making sure that the unit testing and user guide are updated to match.
• Changes must adhere to the CI style guide.
• Each converted class, together with its unit testing and documentation changes, should be handled as a separate PR, referencing this issue.

I will create a feature branch for this, so that it doesn't block the main branch.

Affected classes/helpers:

• system/core/Common (contains the show_error function cited above)
• system/core/Config
• system/core/Lang
• system/core/Loader
• system/core/Router
• system/core/Security
• system/core/URI
• system/database/DB
• system/database/DB_Forge
• system/database/DB_query_builder
• system/database/DB_utility
• system/database/drivers/pdo/pdo_driver
• system/helpers/path_helper
• system/libraries/Driver
• system/libraries/Encryption
• system/libraries/Ftp (also uses lang->line)
• system/libraries/Migration (also uses lang->line)
• system/libraries/Pagination
• system/libraries/Xmlrpc
• system/libraries/Xmlrpcs

[narfbg]

In relation to #4200 ... let's leave show_error() as is for now.

Changing it to just throw an exception may not be as good as I initially thought. It may still happen, but after everything else is done.

[narfbg]

Final: show_error() remains unchanged and a few such calls may remain ... The one triggered by CSRF stays for sure. See #4229.

[narfbg]

@DavidOtis Look above and you'll see 4 pages of mentions that nobody cares about ... Please reference this issue only from your PR descriptions (one reason to actually write a description instead of leaving them empty) and not in every commit message.