fix: handle fractional day conversion for hour-based detection periods (#914)

When detection_period uses period: hour with small counts (e.g., 2),
the conversion to days produces fractional results (2/24 = 0.083).
Previously, casting to int yielded 0, which then failed validation
because the check used "if not value" (treating 0 as falsy).

Changes:
1. Use ceil() in convert_period to round up fractional days, ensuring
   at least 1 day of backfill data when using hourly periods
2. Fix validation logic to check "is none" instead of "not value",
   properly distinguishing between missing (None) and zero (0) values

This ensures that configurations like:
  detection_period: {period: hour, count: 2}
now correctly convert to 1 day instead of 0, and that explicit
zero values are accepted as valid if provided.

Fixes issue where volume_anomalies test with hourly detection periods
would incorrectly report "Missing mandatory configuration: ['backfill_days']"

Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: joostboon

macros/edr/tests/test_configuration/get_anomalies_test_configuration.sql

@@ -99,7 +99,7 @@
     {%- endif %}
 
     {%- for mandatory_param in mandatory_configuration %}
-        {%- if not test_configuration.get(mandatory_param) %}
+        {%- if test_configuration.get(mandatory_param) is none %}
             {%- do missing_mandatory_params.append(mandatory_param) -%}
         {%- endif %}
     {%- endfor %}

macros/edr/tests/test_configuration/get_period_vars.sql

@@ -83,6 +83,6 @@
     {% endif %}
   {% endif %}
 
-  {% set period_count = period_count | int %}
+  {% set period_count = (period_count | round(0, 'ceil')) | int %}
   {{ return({'period': convert_to, 'count': period_count}) }}
 {% endmacro %}