fix(ble): treat Byte17 != 0 as fitting-ok (matches firmware semantics)

Firmware reports fitting via 'Byte17 != 0', not a dedicated bit — during scan
Byte17 carries the wave marker (0x20/0x40/0x60/0x80), all non-zero, so it also
signals fitting-ok. The old 'byte17 & 0x1F == 1' test never matched (low 5 bits
are 0 when the wave marker occupies the high 3), so auto-scan waited forever on
fitting==1 and then mis-fired '未贴合失败'. Verified against real-device frames:
scan now collects and buckets all four wavelengths.

miniprogram/services/ble/protocol.js

@@ -251,7 +251,9 @@ function parseVendorStatus(buffer) {
     for (var p = 0; p < 7; p++) {
       pds19.push(bytes[p * 2] | (bytes[p * 2 + 1] << 8))
     }
-    // Byte17 复用：低位=贴合状态(0/1)，高三位=扫描时光色指示（0 表示无标识/非扫描态或旧固件）。
+    // Byte17 复用：固件约定「非 0 即贴合通过」；高三位另表扫描光色（001红/010红外/011紫外/100暖黄）。
+    //   扫描态 Byte17 = 0x20/0x40/0x60/0x80（有光色→非0→贴合）；检测启动但未贴合时 = 0。
+    //   注意：贴合与光色共用一字节，固件当前无独立贴合位（已反馈，待其拆分）。
     // Byte18 运行状态：0=IDLE 1=检测 2=治疗 0xFF=故障。
     var scanWaveCode = (bytes[16] >> 5) & 0x07
     return {
@@ -259,7 +261,7 @@ function parseVendorStatus(buffer) {
       is_heartbeat: false,
       pd: pds19,
       vbat: bytes[14] | (bytes[15] << 8),
-      fitting: bytes[16] & 0x1F,
+      fitting: bytes[16] !== 0 ? 1 : 0,
       scan_wave: SCAN_WAVE_KEY[scanWaveCode] || null,
       run_state: bytes[17],
       is_fault: bytes[17] === 0xFF,