fix(processing-chain): cast processors to plain dict at entry

build_processing_chain stored values via 'processors[key] = node'. When
processors is a dbetto AttrsDict (which CLI configs are, after
Props.read_from), __setitem__ wraps any plain dict in a fresh
AttrsDict via __init__'s copy loop, decoupling the stored entry from
the local reference. Later in-place mutations like node["prereqs"] = ...
then never reached the entry in processors[key], so
resolve_dependencies failed with KeyError: 'prereqs'.

Cast processors to a plain dict once at the top of the function so all
subsequent assignments behave normally, regardless of the input type.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: gipert

src/dspeed/processing_chain.py

@@ -2372,6 +2372,12 @@ def build_processing_chain(
     if "processors" in processors:
         processors = processors["processors"]
 
+    # cast to a plain dict so that subsequent processors[key] = ... does not
+    # go through wrapping __setitem__ overrides (e.g. dbetto.AttrsDict copies
+    # plain dicts into fresh AttrsDicts on assignment, which would silently
+    # decouple the stored value from any local reference)
+    processors = dict(processors)
+
     buffer_len = len(tb_in) if tb_in is not None else 1
     proc_chain = ProcessingChain(block_width, buffer_len)