improved `mean_insert_size` metric definition

[nicolas-bertin]

proposal to improve mean_insert_size metric definition from

https://github.com/ga4gh/quality-control-wgs/blob/e692682078a3f47b8160cc1ef74614227264b847/metrics_definitions/metrics_definitions.md?plain=1#L83-L88

to [Edit]

https://github.com/ga4gh/quality-control-wgs/blob/bddde198a26f1311a0188f8792b71d6fe704949d/metrics_definitions/metrics_definitions.md?plain=1#L59-L65

see #7

[mhebrard]

• Remove properly paired criteria as popular tools (samtools & picard) do not enforce this filter
• Clarify MQ criteria - include all mapped reads - no MQ filter applied

https://github.com/ga4gh/quality-control-wgs/blob/bddde198a26f1311a0188f8792b71d6fe704949d/metrics_definitions/metrics_definitions.md?plain=1#L59-L65

[justinjj24]

In modifying the metric mean_insert_size computing by using samtools stats it appears that only marginal difference from removing duplicate reads of the 30x good quality data.

However this removing duplicate reads imply the need to run 2 times samtools stats( with & without duplicates) for the metrics pct_reads_mapped and pct_reads_properly_paired heavy compute / marginal gain. Also, a concern about the poor quality data with higher number of duplicate reads!

[mhebrard]

summary of sugegstions:

• Change reference implementation to use samtools stats instead of picard reference implementation already compute samtools stats to gather other metrics
• Keep duplicates

[mhebrard]

• Change reference implementation to use samtools stats
• Keep duplicates

https://github.com/ga4gh/quality-control-wgs/blob/bddde198a26f1311a0188f8792b71d6fe704949d/metrics_definitions/metrics_definitions.md?plain=1#L59-L65