fwrite vs ncputs

theoretically, fwrite() could get better performance than
fputs() due to knowing the length of its input buffer. i
didn't see any perf change beyond noise, but who knows.

src/lib/internal.h

@@ -58,6 +58,7 @@ typedef enum {
 // protocol, we just have to rewrite them.
 typedef struct sprixel {
   char* glyph;       // glyph; can be quite large
+  int glyphlen;
   uint32_t id;       // embedded into glusters field of nccell, 24 bits
   struct ncplane* n; // associated ncplane
   sprixel_e invalidated;

src/lib/render.c

@@ -723,7 +723,7 @@ term_esc_rgb(FILE* out, bool foreground, unsigned r, unsigned g, unsigned b){
   }
   rgbbuf[offset++] = 'm';
   rgbbuf[offset] = '\0';
-  if(ncfputs(rgbbuf, out) == EOF){
+  if(fwrite(rgbbuf, offset, 1, out) != 1){
     return -1;
   }
   return 0;
@@ -971,7 +971,7 @@ rasterize_sprixels(notcurses* nc, const ncpile* p, FILE* out){
       if(goto_location(nc, out, y + nc->stdplane->absy, x + nc->stdplane->absx)){
         return -1;
       }
-      if(ncfputs(s->glyph, out) < 0){
+      if(fwrite(s->glyph, s->glyphlen, 1, out) != 1){
         return -1;
       }
       s->invalidated = SPRIXEL_NOCHANGE;

src/lib/sprite.c

@@ -39,6 +39,7 @@ sprixel* sprixel_create(ncplane* n, const char* s, int bytes, int placey, int pl
       free(ret);
       return NULL;
     }
+    ret->glyphlen = bytes;
     ret->tacache = tacache;
     ret->invalidated = SPRIXEL_INVALIDATED;
     ret->n = n;