msvcrt: exponential notation precision in scanf

The problem is in parsing exponential notation precision lack. The reason is using float instead of double as scaling value and iterative multiplication of source value instead of single one, which leads to error accumulation. Patch is in attachment.

According to testman:
scanf.c:116: Test failed: Got 1.100000491738425600e-030, expected 1.100000000000000000e-030

at:
34 double double_res;
...
113 strcpy(buffer, "1.1e-30");
114 ret = sscanf(buffer, "%lf", &double_res);
116 ok(double_res >= 1.1e-30-1e-45 && double_res <= 1.1e-30+1e-45,
117 "Got %.18le, expected %.18le\n", double_res, 1.1e-30);

After applying patch test is passed successfully.

PS. Is there any sense of fixing those problems exposed by tests, or it is better just to wait for sync with wine?