Add JSON output format and a separate interpreter binary

deck-engine gains --format json (language-independent, slug-based)
so a new dist/interpreter-cli can score how significant a day's
transits are without linking the engine itself — it depends only on
the JSON shape, via its own minimal parser. Also documents the full
reading pipeline end to end (CLAUDE.md, README, docs/) and adds
docs/reading.md, a non-technical explanation of what a daily reading
contains and means.

interpreter/src/reading_io.h

@@ -0,0 +1,23 @@
+#ifndef DECK_READING_IO_H
+#define DECK_READING_IO_H
+
+#include <stdbool.h>
+#include <stddef.h>
+
+/* Header-only dependency on the engine, same policy as significance.h -
+ * see its comment for why. */
+#include "../../engine/src/reading.h"
+
+/* Parses deck-engine's `--format json` output (text, null-terminated at
+ * text[length]) and fills *out with just enough of a DailyReading for
+ * interpret_daily_reading() to work on: out->transits.aspects[]/
+ * aspect_count. out->natal and out->spread are left zeroed - nothing
+ * here reads them (see significance.c), so there's no reason to parse
+ * the rest of the document yet.
+ *
+ * Returns false on malformed JSON, or if "transits"."aspects" isn't
+ * present as an array (an empty array is fine - that's a real "no
+ * aspects today" reading, not an error). */
+bool reading_load_json(const char *text, size_t length, DailyReading *out);
+
+#endif