Performance improvement parsing numbers

[Tangerino]

I made a test in function parse_number.
I replaced the 'strtod' function by the one in sqlite3 project 'sqlite3AtoF' (V3.22.0) and got the parsing process 20%~25% faster in general using my regular JSON documents.
I'm sorry but I can't now do the pull request, if someone is interesting investigate this would be awesome for the project.
Thanks

[FSMaxB]

The speed depends on you C standard libraries implementation of strtod. I looked at sqlite3AtoF and I think I don't really want that kind of complexity in cJSON just for parsing numbers. Also I didn't really look at what they are doing exactly, but their code doesn't seem to be ANSI C, they have a 64bit integer type for example. In cJSON I can't have that because the ANSI C standard doesn't provide one.

[Tangerino]

Understood.
We do have different uses cases. So we may have hook functions for parsing & printing too.
Also a simple code change yields to a better print performance in general. You may then change a bit the print_number function and add a special case where we are dealing with integer numbers as in:

        /* This checks for NaN and Infinity */
        if ((d * 0) != 0) {
            length = sprintf((char*) number_buffer, "null");
        } else {
            // integer number? Avoid unnecessary floating point operations
            int id = (int) d;
            if ((id == d) && (d > INT_MIN) && (d < INT_MAX)) {
                length = sprintf((char*) number_buffer, "%d", id);
            } else {
                /* Try 15 decimal places of precision to avoid nonsignificant nonzero digits */
                length = sprintf((char*) number_buffer, "%1.15g", d);
                /* Check whether the original double can be recovered */
                if ((sscanf((char*) number_buffer, "%lg", &test) != 1) || ((double) test != d)) {
                    /* If not, print with 17 decimal places of precision */
                    length = sprintf((char*) number_buffer, "%1.17g", d);
                }
            }
        }

[FSMaxB]

Detecting integers in print_number is something that I could accept as a performance improvement, if you say that it is substantial.

[FSMaxB]

About adding hooks: I'm not sure how far I am willing to go. But I'll definitely consider it. Although I'm not sure when I will have the time for feature work again (so I can continue with #186).

[FSMaxB]

I commited the integer printing improvement to develop 85f76ba.