[Aikido] Fix shell injection bypass in nested quote context parsing

[aikido-autofix]

This patch addresses a shell injection vulnerability where nested quote contexts could bypass detection by examining only adjacent characters around user input. The fix implements proper shell quoting state parsing that tracks quote context from the beginning of the command rather than relying on local character analysis. Changes were made to is_safely_encapsulated.py to enhance quote context tracking logic, with corresponding test cases added to detect_shell_injection_test.py and is_safely_encapsulated_test.py to validate the fix against nested quote scenarios.

✅ 1 issue fixed by this PR

Issue	Severity	Description
CodeAudit#811707943	HIGH	The helper decides safety from the characters adjacent to the user-controlled substring rather than from shell parsing state. For user_input = "$(id)" and command = 'echo "\'$(id)\'"', it observes matching apostrophes and returns True. Those apostrophes occur inside an active outer double-quoted region and do not disable command substitution. detect_shell_injection() therefore returns before calling contains_shell_syntax(), which would flag $. The false negative affects the shared detection path used before os.system and subprocess.Popen(..., shell=True), allowing a request-controlled value already interpolated into such a command to execute with the consuming process's privileges.