Add test cases for altered_text_cause

[arash77]

Follows up on your review point: altered_text_cause() had no test cases.

You were right. The function is called once, in split_text.py, and nothing
exercised it. It picks between two causes for the same symptom — a chunk whose
text no longer occurs in the input — and the two causes have opposite remedies,
so picking the wrong one sends the user in a circle. That choice was untested.

What is added

Two test cases, one per branch, plus a small input file for each:

• multibyte_persian.txt with the token splitter. gpt2 cuts Persian into
  very small tokens, so a cut lands inside a multi-byte character and the chunk
  comes back with the replacement character. Expected cause: the token cut.
• repeated_separator.txt with a discarded comma. The run of separators is
  rebuilt as a single one, so the chunk text no longer occurs in the input.
  Expected cause: the separator.

Each test asserts the cause it expects and that the other cause is absent,
so swapping the two branches makes both tests fail rather than one.

Notes

• Only tests and test data. No change to the tool logic, so nothing here can
  change the tool's behaviour.
• The function has a third, fallback cause. I tried 96 combinations across the
  four splitters, both keep_separator values, four chunk sizes and four input
  shapes, and it never fired. It looks like a safety net rather than a case that
  can be reached, so I left it uncovered instead of contriving an input for it.

planemo lint is clean and both new tests pass locally (31 tests total, was 29).