GeezerVison

This change does not change the code in any way.
The change formats all tests and some demos according to PEP-8.
Additionally I fixed some missing imports of stackles.test_*-functions,
moved imports to the top of the file, removed some duplicate imports
and added some magic comments to silence my IDE (Eclipse Pydev).

https://bitbucket.org/stackless-dev/stackless/issues/82
(grafted from 43fcfbe017a1456c7285819eee6a2c4984e20045)

Per W25 spec §3 + supervisor [chat L2269] sequencing + theologian
[chat L2268] (1) — extends Step B beyond §1b TUs to close the FULL
drift surface (simplify_c.c shares the §1b drift mechanism but
doesn't include hir_basic_block_c.h pre-W25, hence outside §1b set).

PER-EXTERN CLASSIFICATION (per spec §3 4-class taxonomy):
  Class A (in hir_c_api.h, deleted): 21 unique
    hir_func_alloc_register, hir_c_insert_before, hir_cfg_alloc_block,
    hir_cfg_split_after, hir_c_create_branch_cpp, hir_c_create_cond_branch_cpp,
    hir_func_env_allocate_type_method_cache, hir_func_env_allocate_type_attr_cache,
    hir_c_set_suppress_exc_deopt, hir_c_create_fill_type_method_cache,
    hir_c_create_load_method_cached, hir_c_create_load_module_method_cached,
    hir_c_create_fill_type_attr_cache, hir_c_create_load_attr_cached,
    hir_c_create_load_module_attr_cached, hir_instr_unlink, hir_instr_destroy,
    hir_c_cond_branch_true_bb, hir_c_cond_branch_false_bb,
    hir_cfg_blocks_first_ptr, hir_cfg_blocks_next_ptr
  Class C2 BB (in hir_basic_block_c.h, deleted): 5
    hir_bb_append_instr, hir_bb_first_instr, hir_bb_next_instr,
    hir_bb_prev_instr, hir_bb_remove_phi_predecessor
  Class C1 (NONE): all 26 unique functions in canonical headers post-B-3.5
    (5 of these were promoted in B-3.5: hir_cfg_blocks_first_ptr/next_ptr,
    hir_c_cond_branch_true_bb/false_bb, hir_bb_remove_phi_predecessor —
    transitively closes simplify_c.c drift via B-3.5)

CHANGES (+15/-41):
  Added 2 includes (hir_c_api.h + hir_basic_block_c.h)
  Deleted 29 extern declaration lines (24 unique + 5 dups)
  Added 5 in-place comments noting B-5 cleanup + non-lint extern rationale
  Non-lint externs kept: hir_register_type, hir_output_type,
    hir_phi_create_2way, hir_liveness_is_last_use, hir_assign_create,
    hir_copy_propagation_run, hir_reflow_types_c, hir_clean_cfg_run

VERIFICATION: simplify_c.c now has 0 lint-pattern externs.
Step C lint gate (when shipped) will pass for this TU.

CUMULATIVE B-0..B-5: 7/7 §1b TUs FULLY cleaned + simplify_c.c (post-§1b
addition) = 65 lint externs cleaned across 7 TUs. Drift surface closed.

NOTE: simplify_c.c becomes a §1b TU (now includes hir_basic_block_c.h).
count_w25_b1b_tus.sh next run will show updated 17→18 §1b TUs (or 17 if
the "with extern decls" filter also moves simplify_c.c from no-1b-set
to 1b-with-zero-externs). Re-baseline post-B-5 if needed.

Remaining: B-4 (builder_emit_c.c, 130 externs full-paste discipline) +
Step C lint gate.

testkeeper STRICT verify (gate items 1-3 hardened + verbatim trailer
per CLAUDE.md ba00817) requested post-commit.

Authorization chain:
  - W25 spec §3 + 4-class taxonomy: theologian L2017 + L2189
  - B-5 simplify_c.c addition: theologian L2268 (1) + supervisor L2269
  - In-session adoption: pythia python#82 substance closure