fix(tools): remove shell-based ranged file reads

[failuresmith]

• Closes: ReadFileTool shells out on ranged reads and interpolates untrusted path #5530

Problem:
ReadFileTool uses a shell-based cat -n ... | sed -n ... path for ranged reads when start_line > 1 or end_line is provided. PR #5537 improves the original report by quoting path, but the ranged-read path still shells out and still interpolates end_line into the sed command.

Because tool arguments are not enforced as runtime-validated integers in this path, that leaves a remaining command-injection surface. The shell path also masks missing-file errors for ranged reads by returning ok with empty content.

Solution:
Remove the shell-based ranged-read branch entirely and route all reads through self._environment.read_file(path) plus the existing Python line-slicing logic.

This fixes the issue by eliminating shell interpretation from ReadFileTool rather than trying to quote individual inputs. It also preserves correct missing-file behavior for ranged reads and adds regression coverage for:

• normal ranged reads
• missing-file ranged reads
• non-integer end_line input

Testing Plan

Verified with targeted unit tests for ranged reads and with a manual PoC that previously triggered command execution via end_line.

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

Manual End-to-End (E2E) Tests:

1. Run the adjusted PoC that passes a malicious end_line value to ReadFileTool.
2. Before this change, the marker file is created as a side effect.
3. After this change, ReadFileTool returns an error for non-integer end_line and the marker file is not created.

Observed after the fix:

{'status': 'error', 'error': '`end_line` must be an integer if provided.'}
marker exists: False

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

This is intended as a complete fix for the ranged-read injection surface, rather than a partial mitigation limited to quoting path.

[rohityan]

Hi @failuresmith, Thank you for your contribution through this pull request! This PR has merge conflicts that require changes from your end. Could you please rebase your branch with the latest main branch to address these? Once this is complete, please let us know so we can proceed with the review.