Improve error handling for missing source files

[kcornw]

Raise InvalidSourceList error if new source file is not found. dmypy now correctly reflects the absence of non-existent files, which was previously not reported.

Fixes #18111

---

Testing

I manually tested the changes and confirmed that dmypy now correctly reports errors when attempting to check non-existent files:

$ dmypy run -- a.py
Success: no issues found in 1 source file
$ dmypy run -- nonexistent.py
mypy: can't read file 'nonexistent.py': No such file or directory

I haven't added tests yet, as I'm still familiarizing myself with the testing framework used in the project. If this change is deemed appropriate, I will add the necessary test cases soon to ensure it is thoroughly validated.

I’m eager to contribute further and ensure this fix is robust. Thank you for your support and feedback!

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅

[KotlinIsland]

a test would be good here

[kcornw]

Based on the existing test cases, particularly (in test-data/unit/fine-grained.test):

[case testNonExistentFileOnCommandLine4]
# cmd: mypy a.py nonexistent.py
[file a.py]
[file nonexistent.py]
[delete nonexistent.py.2]
[out]
==

It appears that the behavior of ignoring a nonexistent file after caching the previous result is intentional? (Please let me know if my understanding of this test case is correct.) This could be why some tests are currently failing.

Additionally, if the change proposed in this PR is appropriate, we may need to modify many existing test data to align with the new behavior.