1 - You can REPLACE _x000D_ with nothing to remove it

2 - You could also REPLACE P, or use SUBSTR to return all the characters from the second onward.

To make customer_id a number, you'll need to add a column to do this:

create table t (
  cust_id varchar2(10),
  text    varchar2(1000)
);

insert into t 
  values ( 'P112113', 'This is a sentence_x000D_. This is another_x000D_.' );
  
alter table t
  add cust_id_num integer;
  
update t
set    text = replace ( text, '_x000D_' ),
       cust_id_num = substr ( cust_id, 2 );
       
select * from t;

CUST_ID   TEXT                                   CUST_ID_NUM   
P112113   This is a sentence. This is another.        112113 

Or - as you're changing most (all?) the rows in the table, you could use the create-table-as-select trick to update the data, then switch the tables over:

rollback;

select * from t;

CUST_ID   TEXT                                                 CUST_ID_NUM   
P112113   This is a sentence_x000D_. This is another_x000D_.        <null> 

create table tmp as 
  select cast ( substr ( cust_id, 2 ) as int ) cust_id, 
         replace ( text, '_x000D_' ) text
  from   t;
  
select * from tmp;

CUST_ID  TEXT                                   
 112113  This is a sentence. This is another.  

drop table t;
rename tmp to t;

select * from t;

CUST_ID  TEXT                                   
 112113  This is a sentence. This is another.  

This will be faster than UPDATE, but on 15,000 rows there will probably be little difference. You'll have to copy indexes, constraints, etc. from the old table to the new too. If this is a live application, you can do this online with dbms_redefinition.