fix: make streaming reasoning-tag filter case-insensitive

The streaming think-tag suppressors in cli.py (_stream_delta) and
gateway/stream_consumer.py (_filter_and_accumulate) matched tag names
with case-sensitive str.find(), so only the exact-case literals in the
tag tuples were caught. Mixed-case variants a model may emit — <Think>,
<ThInK>, <REASONING>, <Thought> — slipped through and leaked raw
reasoning into the user-visible stream.

Match against a lowercased view of the buffer with lowercased tag names
at all three sites (open-tag boundary search, partial-tag hold-back,
close-tag search) in both paths. Only KNOWN tag names are matched — no
substring matching — and the block-boundary gating that protects prose
mentions of <think> is preserved.

- 6 parametrized case-insensitive regression tests in each of
  tests/gateway/test_stream_consumer.py and
  tests/cli/test_stream_delta_think_tag.py.

Salvaged from PR #27289 by @YLChen-007.

cli.py

@@ -5443,10 +5443,14 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
             self._stream_last_was_newline = True  # start of stream = boundary
 
         if not getattr(self, "_in_reasoning_block", False):
+            # Case-insensitive matching against a lowercased view so
+            # mixed-case tag variants (<Think>, <THINKING>, …) are caught.
+            prefilt_lower = self._stream_prefilt.lower()
             for tag in _OPEN_TAGS:
+                tag_lower = tag.lower()
                 search_start = 0
                 while True:
-                    idx = self._stream_prefilt.find(tag, search_start)
+                    idx = prefilt_lower.find(tag_lower, search_start)
                     if idx == -1:
                         break
                     # Check if this is a block boundary position
@@ -5486,11 +5490,12 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
 
             # Could also be a partial open tag at the end — hold it back
             if not getattr(self, "_in_reasoning_block", False):
-                # Check for partial tag match at the end
+                # Check for partial tag match at the end (case-insensitive)
                 safe = self._stream_prefilt
                 for tag in _OPEN_TAGS:
+                    tag_lower = tag.lower()
                     for i in range(1, len(tag)):
-                        if self._stream_prefilt.endswith(tag[:i]):
+                        if prefilt_lower.endswith(tag_lower[:i]):
                             safe = self._stream_prefilt[:-i]
                             break
                 if safe:
@@ -5503,8 +5508,9 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin):
         # Keep accumulating _stream_prefilt because close tags can arrive
         # split across multiple tokens (e.g. "</REASONING_SCRATCH" + "PAD>...").
         if getattr(self, "_in_reasoning_block", False):
+            prefilt_lower = self._stream_prefilt.lower()
             for tag in _CLOSE_TAGS:
-                idx = self._stream_prefilt.find(tag)
+                idx = prefilt_lower.find(tag.lower())
                 if idx != -1:
                     self._in_reasoning_block = False
                     # When show_reasoning is on, route inner content to