Feat: Support for Distributions in Templated Workloads

[ETHenzlere]

PR adds support for various Numerical Distributions in templated benchmarks

• Uniform
• Binomial
• Zipfian
• Scrambled

The distributions work not only for integers, but also different types such as timestamps, long, float
A table of supported combinations can be found in the templated benchmarks readme.

Type	uniform	binomial	zipfian	scrambled (zipfian)
INTEGER	X	X	X	X
FLOAT / REAL	X	X	-	-
BIGINT	X	X	X	X
VARCHAR / STRING	X	-	-	-
TIMESTAMP	X	X	X	X
DATE	X	X	X	X
TIME	X	X	X	X

Usage example:

<templates>
    <template name="MyTemplate">
        <query><![CDATA[SELECT * FROM MyTable WHERE id = ?]]></query>
        <types>
            <type>INTEGER</type>
        </types>
        <values>
            <value dist="uniform" min="0" max="1000" seed="1"/>
        </values>
         <values>
            <value dist="zipf" min="1" max="1000" seed="2"/>
        </values>
    </template>
    <!-- ... -->
<templates>

This PR is not a breaking change - One can still use a static value in the templated queries.
<value>10</value>

In the future, I could see a breaking change that adds the datatype directly to the values so the TemplatedValue can do the type handling directly. <value dist="uniform" type="integer">. This would make type checking easier and remove the need to store the original min/max values as strings for all datatypes.

[bpkroth]

Looking pretty good. Just a few cosmetic polish things left:

• documentation or improved handling of timestamps as longs issue in the configs
• enums instead of static strings
• expanded test coverage for a few cases

[bpkroth]

LGTM. Thanks!