fix(env): Return None in the get_str on unset vars

[lu-zero]

Fixes IFS handling when unset in local scope and potentially makes the HOME fallback more correct.

The get_str method was incorrectly converting unset variables to empty strings instead of returning None. This caused issues with IFS handling where unset IFS variables should fall back to default behavior but were instead treated as empty IFS.

[github-actions]

Performance Benchmark Report

Benchmark name	Baseline (μs)	Test/PR (μs)	Delta (μs)	Delta %
clone_shell_object	16.91 μs	17.19 μs	0.28 μs	🟠 +1.64%
eval_arithmetic	0.16 μs	0.16 μs	-0.00 μs	⚪ Unchanged
expand_one_string	1.56 μs	1.55 μs	-0.01 μs	⚪ Unchanged
for_loop	23.70 μs	24.13 μs	0.44 μs	⚪ Unchanged
full_peg_complex	57.37 μs	57.93 μs	0.56 μs	⚪ Unchanged
full_peg_for_loop	6.26 μs	6.31 μs	0.05 μs	⚪ Unchanged
full_peg_nested_expansions	16.84 μs	17.05 μs	0.21 μs	⚪ Unchanged
full_peg_pipeline	4.18 μs	4.22 μs	0.04 μs	⚪ Unchanged
full_peg_simple	1.75 μs	1.75 μs	-0.00 μs	⚪ Unchanged
function_call	2.19 μs	2.22 μs	0.03 μs	⚪ Unchanged
instantiate_shell	54.66 μs	53.66 μs	-0.99 μs	🟢 -1.82%
instantiate_shell_with_init_scripts	27141.27 μs	27592.17 μs	450.91 μs	⚪ Unchanged
parse_peg_bash_completion	2092.90 μs	2135.59 μs	42.69 μs	🟠 +2.04%
parse_peg_complex	18.04 μs	19.35 μs	1.31 μs	🟠 +7.26%
parse_peg_for_loop	1.90 μs	1.91 μs	0.01 μs	⚪ Unchanged
parse_peg_pipeline	1.99 μs	1.93 μs	-0.06 μs	⚪ Unchanged
parse_peg_simple	1.02 μs	1.04 μs	0.02 μs	⚪ Unchanged
run_echo_builtin_command	15.50 μs	15.84 μs	0.34 μs	⚪ Unchanged
tokenize_sample_script	3.90 μs	3.73 μs	-0.17 μs	🟢 -4.38%

Code Coverage Report: Only Changed Files listed

Package	Base Coverage	New Coverage	Difference
Overall Coverage	🟢 74.38%	🟢 74.38%	⚪ 0%

Minimum allowed coverage is 70%, this run produced 74.38%

Test Summary: bash-completion test suite

Outcome	Count	Percentage
✅ Pass	1558	73.87
❗️ Error	18	0.85
❌ Fail	179	8.49
⏩ Skip	339	16.07
❎ Expected Fail	13	0.62
✔️ Unexpected Pass	2	0.09
📊 Total	2109	100.00

[reubeno]

Thanks for the fix, @lu-zero!

Is this separate from #988, then? Could you please add a YAML-based test case that fails without this fix?

[lu-zero]

we can close this one and keep the other. (sorry for not noticing the reply till now)