ENH: Avoiding conversion to NumPy Scalar. Implementation Copy-paste tag in conv_template.py

[arinkverma]

This pr is based on same rationale and improvement suggested in @raulcota 's pr #2941. For some types, it is a lot faster to just extract the value directly from the Python scalar. It take care of main concerned of code duplication.

Implementation of copy and paste tag in cove_template.py

copy and paste tag

Example
Here is demo template

/**begin copy 
 *@hi@
 */
Hello! This is just a dummy line
 /**end copy**/

First I will write some useless line, then include code pasted to variable 
hi as /**paste @hi@**/

Output:

First I will write some useless line, then include code pasted to variable 
hi as Hello! This is just an dummy line

Usage

Here, we could avoid duplicating the whole function body just to add one specific check as per dtype.

Speed Up in Int

1.6 Array * Array
1.0 PyInt * Array
1.1 NyInt32 * Array
1.0 PyInt + Array
1.1 NyInt32 + Array
1.0 PyInt * PyInt
1.0 NyInt32 * NyInt32
2.2 PyInt * NyInt32
2.0 PyInt * Array[1]
2.2 PyInt + NyInt32
8.2 PyInt < NyInt32
1.0 NyInt32 * NyInt32 * NyInt32 * NyInt32 * NyInt32
2.2 NyInt32 ** 2.1828018936
0.9 PyInt ** 2