test/docs: close the env_int64 ban hole, drop a require the migration orphaned

Copilot review, #3580, two findings, both real:

BANNED_HELPERS listed env_or / pf_env_int / win_env_int / env_int / env_str /
env_flag but NOT env_int64 — while the comment right above it names env_int64
as one of the five shapes the tree grew, and the failure message tells you to
call it. A re-added local env_int64 would have slipped through and taken its
knob out of the coverage scan with it. Verified by re-adding one: the test now
names the file and the signature.

dasllama_math's  carried a comment about to_int in the
DASLLAMA_CALLER_PRIO override. That path is env_int now — and the require is
not merely mis-commented, it is dead: the engine compiles clean without it.
Removed rather than re-worded, since a correct comment on an unused require is
still an unused require.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: borisbat

modules/dasLLAMA/dasllama/dasllama_math.das

@@ -4,7 +4,6 @@ options _comment_hygiene = true
 module dasllama_math shared public
 
 require math
-require strings  // to_int: DASLLAMA_CALLER_PRIO env override in setup_dasllama_jobque_
 require daslib/fio  // has_env_variable: DAS_JOBQUE_TEAM_RANK_GATE suppresses the profile's team_rank_gate knob
 require dasllama/dasllama_env  // env_flag/env_int64: the shared typed readers + the knob registry
 require daslib/jobque_boost public  // matmul's threaded path expands parallel_for; consumers need its symbols

modules/dasLLAMA/tests/test_env_registry.das

@@ -116,6 +116,8 @@ let BANNED_HELPERS = [
     "def env_or(",
     "def private pf_env_int(",
     "def private win_env_int(",
+    "def env_int64(",
+    "def private env_int64(",
     "def env_int(",
     "def private env_int(",
     "def env_str(",