feat(types): pass through upstream cache tokens in TokenUsage

dsh displays cache-hit %, but llmsproxy dropped every upstream's cache
fields — deepseek prompt_cache_hit_tokens, OpenAI prompt_tokens_details.
cached_tokens, anthropic cache_read_input_tokens, gemini cachedContentTokenCount.

Changes:
- TokenUsage: add PromptTokensDetails (with CachedTokens) + PromptCacheHit/Miss
- MarshalJSON: emit prompt_tokens_details.cached_tokens (OpenAI v2 standard)
  and prompt_cache_hit/miss_tokens (DeepSeek legacy) — dsh reads the former
  first, falls back to the latter
- mergeUsage: preserve cache fields across stream chunks
- standardSSEChunk: parse the upstream raw prompt_tokens_details too
- deepseek.lua: forward prompt_cache_hit/miss_tokens + create
  prompt_tokens_details from them
- openai.lua: forward prompt_tokens_details.cached_tokens and legacy
  prompt_cache_hit/miss_tokens; normalize legacy hits into the standard
  object so dsh sees them regardless of upstream format
- anthropic.lua: map cache_read_input_tokens → prompt_tokens_details
- gemini.lua: map cachedContentTokenCount → prompt_tokens_details

internal/types/types.go

@@ -84,6 +84,22 @@ type TokenUsage struct {
 	Prompt     int `json:"prompt"`
 	Completion int `json:"completion"`
 	Total      int `json:"total"`
+	// PromptTokensDetails mirrors the OpenAI v2 usage.prompt_tokens_details
+	// object so cache-hit counts reported by OpenAI-compatible upstreams
+	// (and by adapters that normalize their own cache fields into it) pass
+	// through to clients that read it — dsh reads cached_tokens from here.
+	PromptTokensDetails *PromptTokensDetails `json:"prompt_tokens_details,omitempty"`
+	// PromptCacheHit / PromptCacheMiss carry the DeepSeek-legacy standalone
+	// fields; dsh falls back to prompt_cache_hit_tokens when
+	// prompt_tokens_details.cached_tokens is absent.
+	PromptCacheHit  int `json:"prompt_cache_hit_tokens,omitempty"`
+	PromptCacheMiss int `json:"prompt_cache_miss_tokens,omitempty"`
+}
+
+// PromptTokensDetails is the OpenAI v2 prompt_tokens_details object. Only
+// CachedTokens is emitted (omitempty drops the whole object when zero).
+type PromptTokensDetails struct {
+	CachedTokens int `json:"cached_tokens,omitempty"`
 }
 
 // MarshalJSON emits both the legacy short keys (prompt/completion/total, used
@@ -92,19 +108,25 @@ type TokenUsage struct {
 // clients such as DSH and DevEco Code read the *_tokens fields.
 func (t TokenUsage) MarshalJSON() ([]byte, error) {
 	return json.Marshal(struct {
-		PromptTokens     int `json:"prompt_tokens"`
-		CompletionTokens int `json:"completion_tokens"`
-		TotalTokens      int `json:"total_tokens"`
-		Prompt           int `json:"prompt"`
-		Completion       int `json:"completion"`
-		Total            int `json:"total"`
+		PromptTokens          int                   `json:"prompt_tokens"`
+		CompletionTokens       int                   `json:"completion_tokens"`
+		TotalTokens            int                   `json:"total_tokens"`
+		Prompt                 int                   `json:"prompt"`
+		Completion             int                   `json:"completion"`
+		Total                  int                   `json:"total"`
+		PromptTokensDetails    *PromptTokensDetails  `json:"prompt_tokens_details,omitempty"`
+		PromptCacheHitTokens   int                   `json:"prompt_cache_hit_tokens,omitempty"`
+		PromptCacheMissTokens  int                   `json:"prompt_cache_miss_tokens,omitempty"`
 	}{
-		PromptTokens:     t.Prompt,
-		CompletionTokens: t.Completion,
-		TotalTokens:      t.Total,
-		Prompt:           t.Prompt,
-		Completion:       t.Completion,
-		Total:            t.Total,
+		PromptTokens:         t.Prompt,
+		CompletionTokens:     t.Completion,
+		TotalTokens:          t.Total,
+		Prompt:               t.Prompt,
+		Completion:           t.Completion,
+		Total:                t.Total,
+		PromptTokensDetails:  t.PromptTokensDetails,
+		PromptCacheHitTokens:  t.PromptCacheHit,
+		PromptCacheMissTokens: t.PromptCacheMiss,
 	})
 }