Add TCGA RNA-seq samples to TPM boxplots

[logstar]

Add TCGA RNA-seq samples in OpenPedCan-analysis data release to TPM boxplots. TCGA RNA-seq sample TPM values are in tcga-gene-expression-rsem-tpm-collapsed.rds, and the sample metadata are in histologies.tsv.

The following design issues have been resolved, but they are still open to suggestions, according to discussions with @jharenza and @chinwallaa in https://opentargetspediatrics.slack.com/archives/C03AR8ZD0RF:

• Show all TCGA samples as additional boxes on the right side of the boxplots.
• Each TCGA box represents a cancer_group in histologies.tsv. TCGA boxes do not need to be limited to the cancer_groups of the pediatric tumor boxes shown in the boxplot, because TCGA cancer_groups do not match 1:1 to pediatric tumor cancer_groups, according to @jharenza's comment at https://opentargetspediatrics.slack.com/archives/C03AR8ZD0RF/p1653496395764939?thread_ts=1653494629.301359&cid=C03AR8ZD0RF.
• Each TCGA box includes all independent samples, because most TCGA cancer_groups only have primary tumor samples, as suggested by @jharenza at https://opentargetspediatrics.slack.com/archives/C03AR8ZD0RF/p1653497428257809?thread_ts=1653494629.301359&cid=C03AR8ZD0RF. The tumor_descriptors of TCGA samples are checked at https://opentargetspediatrics.slack.com/archives/C03AR8ZD0RF/p1653510453690229?thread_ts=1653494629.301359&cid=C03AR8ZD0RF.
• The TCGA boxes take a different color from pediatric tumor boxes or GTEx tissue boxes, as suggested by @chinwallaa.
• Add a parameter to switch include/exclude/collapse TCGA samples, if the boxplot code complexity is still going to be tractable.

cc @taylordm