Add Classed error conditions for programmatic error handling

[Mukulyadav2004]

Fixes #5913
This PR implements 4 strategic classed error conditions to enable programmatic error handling in data.table, these changes add specific error classes to the most common user-reported errors.

Problem:
Currently, data.table errors can only be caught generically using tryCatch(..., error = function(e) ...). This makes it difficult for users to:

• Handle specific error types programmatically: Cannot distinguish between different failure modes.
• Have to parse error messages manually to determine error type.

Solution:
Starting implementing with some of the high impact error classes

• dt_missing_column_error - Missing column errors during key setting, joins, and subsetting
• dt_unsortable_type_error - Unsortable types in key operations
• dt_join_type_mismatch_error - Type incompatibility in joins operations
• dt_invalid_input_error - Input validation failures that break data pipelines

Now the plan is to add more classes based on user feedback.

@MichaelChirico @tdhock @jangorecki can you please review

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.77%. Comparing base (c446f2e) to head (5be1427).
Report is 24 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #7139      +/-   ##
==========================================
+ Coverage   98.51%   98.77%   +0.26%     
==========================================
  Files          81       81              
  Lines       15013    15197     +184     
==========================================
+ Hits        14790    15011     +221     
+ Misses        223      186      -37     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

• HEAD=issue#5913 stopped early for fread disk overhead improved in #6925
  

Generated via commit 5be1427

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	2 minutes and 46 seconds
Installing different package versions	39 seconds
Running and plotting the test cases	2 minutes and 47 seconds

[MichaelChirico]

Thanks, this is a great start! cc @tdhock and @jangorecki, it will be good to have extra eyes here :)

[tdhock]

the names are reasonable.
can you please add a test and/or documentation in a vignette?

[aitap]

This PR implements 5 strategic classed error conditions
<...>

• dt_missing_argument_error - Missing required arguments

Is this PR supposed to contain the dt_missing_argument_error class?

[Mukulyadav2004]

yes, it was supposed to contain the dt_missing_argument_error class. I was going to use this in stopf("A non-empty vector of column names is required for by.x and by.y.") but changed it as it was more like invalid_input

[Mukulyadav2004]

I've added the script for checking error type condition class as of now. Can you please review.

[jangorecki]

While this reference is not bad, it unnecessarily uses third party packages. If there is an alternative that uses base R packages only, then I would prefer to use that one.

[Mukulyadav2004]

Would using ?warning and ?signalCondition be appropriate here?

[jangorecki]

Haven't read them.
A decent stackoverflow topic. We have an internal helper for exception handling in data.table:

data.table/R/test.data.table.R

Line 435 in 2f0d12f

# Thanks to tryCatch2 by Jan here : https://github.com/jangorecki/logR/blob/master/R/logR.R#L21

[MichaelChirico]

It's a good idea to link our ?test, yes. I still think we'd be remiss to omit the Advanced R chapter, which is by now a standard reference here. We can add others too.

[aitap]

What about the first edition of Advanced R? There are mentions of pryr, but all exercises can be done using only base R, I think, and none of the code snippets depend on rlang. (Yes, the link is http:, and https: won't work. Since it's not a redirect, R CMD check won't yell at us.)